<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Dehumo Bickersteth on Medium]]></title>
        <description><![CDATA[Stories by Dehumo Bickersteth on Medium]]></description>
        <link>https://medium.com/@dehumo?source=rss-7b135f973c88------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*DnErf1bQ6bkuS5E60P-HJA@2x.jpeg</url>
            <title>Stories by Dehumo Bickersteth on Medium</title>
            <link>https://medium.com/@dehumo?source=rss-7b135f973c88------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 07:21:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@dehumo/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Begin with the Person and the Responsibility]]></title>
            <link>https://medium.com/human-at-work/begin-with-the-person-and-the-responsibility-8cc5d30910ba?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/8cc5d30910ba</guid>
            <category><![CDATA[professional-development]]></category>
            <category><![CDATA[carve]]></category>
            <category><![CDATA[formation]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 00:25:09 GMT</pubDate>
            <atom:updated>2026-09-17T00:25:09.396Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ktpuazCZ_xxAnv4T3k80Ng.png" /></figure><p>Consider an experienced analyst who has been asked to lead a small team. She knows the subject, produces reliable work and is often the person colleagues consult when an analysis becomes difficult. She wants the new responsibility, partly because she would like other people to develop the confidence she has acquired. Within a few weeks, however, she is reviewing almost everything herself and working later to keep the team’s commitments.</p><p>If we describe her need as delegation skills, we have a recognisable subject for a programme. We can explain a model, provide examples and give her somewhere to practise a conversation. All of that may help. But the description leaves several questions unanswered about the person and the responsibility she is trying to carry.</p><p>What does she believe she is answerable for when somebody else produces the analysis? Can she recognise which decisions a colleague is ready to make? Does she know how to establish a standard without prescribing every step? And has the organisation actually given her time and authority to develop the team, or simply added that expectation to her previous workload?</p><p>These questions would change the learning experience I designed. They would also change the kind of conversation I wanted an AI-supported tool to have with her.</p><p>In CARVE, I bring Capabilities, Activities, Results, Value and Evaluation together so we can examine those relationships. We can begin with whichever part of the situation has drawn our attention. In this example, the long hours might be the first concern, but following them into the work helps us establish what a useful change would mean.</p><p>The team needs dependable analysis that supports decisions. Its members also need opportunities to become more capable of producing and judging that analysis. The new manager wants to support them while having a working life she can sustain. Those purposes can inform the results we seek, including reliable delivery, clearer decisions about who handles what and colleagues becoming ready for more demanding work.</p><p>None of those results is established by recording that she has delegated three tasks. We need to understand what she delegated, what support was available and what happened. Delegating a familiar task to someone already competent may improve the distribution of work. Giving a colleague a carefully supported opportunity to examine an unfamiliar problem may serve an additional developmental purpose. Both can be worthwhile, but they ask different things of her.</p><p>This makes the capability inquiry more specific. She may need knowledge about how to judge readiness, skill in explaining the expected result and practice responding when a colleague proposes an approach she would not have chosen. She may also need to examine a belief that a good manager should always have the answer immediately. Her knowledge, skill and will are connected through the decisions she makes in the situation.</p><p>I would be careful about turning will into a polite word for blaming someone. A reluctance to delegate might reflect an assumption that deserves examination. It might also reflect previous experience of being held responsible for mistakes without being allowed to set the conditions for the work. We need to understand that experience before prescribing greater confidence. The developmental response may include a conversation with her own manager about authority, resources and expectations.</p><p>The same care matters when we examine someone’s posture towards AI. She might use it comfortably to explore a personal interest while hesitating to use it with her team. The difference could concern the information involved, uncertainty about permission or the consequences of a poor recommendation. It could also concern how she understands her professional value. We learn more by examining a particular use than by placing her on a general scale from resistant to enthusiastic.</p><p>An initial conversation in a formation tool could therefore invite her to describe one recent occasion when she took work back from a colleague. What result was she trying to protect? What had she noticed? What did she believe would happen if she left the work with them? The tool could help organise her account and identify questions for further examination, while giving her an opportunity to correct its interpretation.</p><p>Suppose she realises that she never made the decision standard clear. That would suggest a practical focus for development. If instead she had explained the standard but the colleague lacked access to essential information, another explanation becomes more plausible. The point of the conversation is to arrive at a better account of the situation, one specific enough to guide action. It should not make every account end with the same recommended course.</p><p>The person also needs room to decide what she wants from the development. Perhaps she discovers that she values developing others but wants to retain a substantial role in analysis. Perhaps she eventually chooses to deepen her expertise without moving further into management. The <a href="https://pulse.euchreston.com/tools/career-catalyst">Whole Ladder</a> keeps those different professional contributions visible. Adult development needs to respect the possibility that growing more capable changes someone’s direction as well as their readiness for the direction already proposed.</p><p>That matters beyond employment. Someone developing an independent practice may be trying to create a service that expresses their expertise in a different way. Someone returning to work may need to reconnect earlier experience with changed expectations. A person learning music may begin with a desire to express something they cannot yet make audible. In each case, a meaningful intention gives us a reason to inquire, but the inquiry can also refine the intention.</p><p><em>My own experience of creating with technology has repeatedly involved that movement. Working on an idea shows me possibilities and difficulties that were not apparent when I began. I may discover that the result I named was too narrow, or that the part I most wanted to create serves a different purpose from the one I first described. A development design should leave room for that learning, including a considered change of goal.</em></p><p>For <a href="https://pulse.euchreston.com/">Pulse</a>, this affects how I think about profiles, instruments and the conversations that follow them. A profile can help the person bring relevant experience into view. An instrument can provide a structured starting account. Neither should settle the person’s identity or turn an unanswered question into a conclusion about their ability. The experience should help them connect what is known with a responsibility or aspiration they recognise as their own.</p><p>Personalisation then becomes a matter of educational judgment. We need to consider which example will help, what practice is appropriate and whether the next useful step is an explanation, an attempt or a conversation with somebody who can change the conditions. Adjusting the language around a standard recommendation is a much smaller response to the information the person has shared.</p><p>For those of us designing adult learning, this inquiry belongs at the beginning and remains available as the experience develops. Our first account will be incomplete. The person’s attempts, questions and choices can show us what needs reconsideration, provided the design allows those contributions to influence what happens next.</p><blockquote>Think about a programme or tool you are developing now. What do you need to understand about the person and the responsibility before you can decide which learning experience would actually help?</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8cc5d30910ba" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/begin-with-the-person-and-the-responsibility-8cc5d30910ba">Begin with the Person and the Responsibility</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Are We Helping Adults Become?]]></title>
            <link>https://medium.com/human-at-work/what-are-we-helping-adults-become-ca856ccd2bd9?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/ca856ccd2bd9</guid>
            <category><![CDATA[edtech]]></category>
            <category><![CDATA[professional-development]]></category>
            <category><![CDATA[ai-and-adult-learning]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Mon, 14 Sep 2026 00:35:57 GMT</pubDate>
            <atom:updated>2026-09-14T00:35:57.938Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J95EPnyt4_G3a5JKHf5zoA.png" /></figure><p>Some of my earliest experiences of learning with technology came from wanting to make music. I could hear something I wanted to play, and the distance between that intention and what I could produce gave me a reason to practice, watch other people, and learn how the equipment worked. A technique became interesting because it helped me do something I cared about.</p><p>Computer-based music production extended that relationship. With a keyboard and virtual instruments, I could build an arrangement across parts I could not have played together on my own. As I explored, I discovered possibilities that changed the music I wanted to make. <em>My intention developed through the attempt, because hearing what I had produced gave me something new to consider.</em></p><p>The technology was participating in my development through what it allowed me to try and examine. I still had to listen, decide whether a part contributed, and recognize when another adjustment was taking the music away from what I wanted it to express. Having more possibilities gave me more to learn how to judge.</p><p>When I began teaching music production, I wanted to help students develop that relationship themselves. Learning the controls mattered because it gave them greater freedom to act on an intention. They also needed opportunities to hear what their choices produced, explain what they were trying to achieve, and reconsider an approach. Sometimes the attempt helped them discover what they wanted to express. Sometimes they needed help bringing an expanding set of possibilities to a conclusion.</p><p>Those questions followed me into instructional design and corporate learning. <em>What was someone trying to become capable of doing? What experience would help them develop that capability? How could technology support their participation, and what would give us grounds for saying the experience had helped?</em></p><p>In L&amp;D, answering those questions often required me to learn about a field beyond my own. Subject-matter experts brought understanding of their work, while I brought understanding of how to design and produce a learning experience. We had to examine the relationship between those contributions. A program could be accurate about a subject and still give people too little opportunity to learn how to act in the situations they would face.</p><p>This is the history I brought to AI. I already knew that technology could extend what I could create and that working with it could develop my understanding. What interested me was the possibility of a technology that could also participate in the conversation about the work.</p><p>I could bring an idea I had not fully resolved, ask for possible approaches, examine a recommendation, and continue into making something. When I encountered an unfamiliar concept, I could ask for an explanation related to the decision in front of me. When a proposed approach seemed wrong, explaining my objection required me to become more precise about what I intended.</p><p>That is what I mean by cognitive partnership. AI could contribute to framing the problem, developing the idea, and producing the result. Alongside it, I began to recognize the possibility of cognitive apprenticeship: learning through participation in work where explanations and decisions became available to examine. I could ask why an approach had been suggested, what an alternative would change, and what evidence we would need before proceeding.</p><p>The developmental value depended on what I did with those contributions. <em>Accepting an explanation could move the immediate task forward. Investigating it could leave me with a distinction or a question I could use again.</em> In unfamiliar territory, I also needed sources, observations or people capable of helping me check what I could not yet judge confidently myself.</p><p>As a learning designer, I became interested in the structure of these exchanges. Some conversations helped me clarify an intention. Others exposed an assumption, connected an unfamiliar idea with something I understood or gave me a way to examine a disappointing result. Their usefulness depended partly on the questions being asked and on how the conversation developed in response to what I contributed.</p><p>I began approaching those conversations as development interventions that could be deliberately designed. That meant considering what the person needed to understand or practise, what they should be invited to explain, when an example would help and when they needed an opportunity to make a judgment of their own. It also meant connecting the exchange with something they could attempt beyond the conversation.</p><p>Consider a manager who keeps postponing feedback to a colleague. They may know the recommended steps and still expect a direct conversation to damage the relationship. A useful intervention could help them describe the situation, examine that expectation and consider what their silence is already doing. They might then explore possible approaches, rehearse how to begin and decide what they need to establish with the colleague.</p><p>The manager’s responses should influence what happens next. Someone who cannot distinguish an observation from an interpretation may need an example. Someone who understands the distinction but fears the consequences of speaking needs a different kind of inquiry. After the real conversation, there is another opportunity to examine what happened and what they would change.</p><p>Designing that experience requires us to hold a developmental purpose while remaining responsive to the person. The sequence matters, as does the reason for asking the next question.<em> A fluent conversation may feel satisfying, but its educational value becomes clearer when we can connect it with a change in understanding, a considered action or a better judgment in a later situation.</em></p><p>CARVE helps me keep those relationships in view. Capabilities, Activities, Results, Value, and Evaluation connect what a person knows, can do, and is willing to attempt with their decisions, actions, and the difference they are trying to make. Evaluation brings experience back into the inquiry. We can examine whether the difficulty concerns understanding, skill, belief, the chosen activity, or conditions that need attention.</p><p>This is part of what I am encoding into the design of formation tools within <a href="https://pulse.euchreston.com/">Pulse</a>. The educational work includes deciding what an interaction should help someone notice, which questions belong together, what material should inform the exchange and how the person can carry it into practice. AI makes the interaction responsive, while the design gives it a purpose and a basis for judging whether it is useful.</p><p>Building those tools also develops the methodology. An instruction such as “help the person reflect” leaves many decisions unresolved. Making it work requires me to specify what they are reflecting on, what evidence they have, which assumptions deserve examination and what they might do with the resulting understanding. Testing the experience can reveal where my own account of the method needs to become clearer.</p><p>That brings me back to music. Creating something gives me a result through which I can examine my thinking. The tool expresses part of my understanding of development, and working on it gives me reasons to revise that understanding. AI participates both in building the experience and in some of the interactions the experience offers.</p><p>The wider <a href="https://pulse.euchreston.com/">Pulse ecosystem</a> follows from the same concern for continuity. A book can give someone language for a problem. A program can organize opportunities to practice. An instrument can help them examine their starting point and return to it later. A tool or structured conversation can help them work through a situation. Coaches, managers, peers and communities contribute observations, relationships and opportunities that give the development consequences beyond the application.</p><p>An adult might enter through any of those experiences. They bring a life, a history of practice, things they want to create and responsibilities they are trying to grow into. Someone preparing to manage others needs different support from someone deepening a craft or reconsidering their career. The connections should help them pursue what matters in their circumstances and return with what they discover.</p><p>This is what I mean by designing for formation. <em>Information changes what you know; formation changes who you become.</em> In adult and professional development, I want that second statement to be concrete enough to examine: what someone now understands, what they can judge, what they are prepared to attempt and how they respond when the situation changes.</p><p>My experience of building <a href="https://pulse.euchreston.com/">Pulse</a> gives me a place to investigate those possibilities. It also gives me an obligation to ask whether the experiences are doing what I intend. A useful output, an engaging conversation and a developing practitioner are related achievements, and each needs evidence appropriate to the claim.</p><p>For those of us working in edtech, the educational decisions reach into the questions our products ask, the choices they leave with the learner and the opportunities they provide to return to experience. <em>Which part of the learning experience you are designing asks an adult to exercise the judgment you hope they will develop?</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ca856ccd2bd9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/what-are-we-helping-adults-become-ca856ccd2bd9">What Are We Helping Adults Become?</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Three Platforms, One Substrate: What Population-Scale Capability Infrastructure Looks Like]]></title>
            <link>https://medium.com/human-at-work/three-platforms-one-substrate-what-population-scale-capability-infrastructure-looks-like-47c1c65cf680?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/47c1c65cf680</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[capability-development]]></category>
            <category><![CDATA[practitioner-intelligence]]></category>
            <category><![CDATA[future-of-work]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Wed, 20 May 2026 05:34:04 GMT</pubDate>
            <atom:updated>2026-05-20T05:34:04.212Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-I1C_P5oUBcOzbU2V8hZXA.png" /></figure><h3>What a single practitioner’s becoming looks like across the arc of a career, on a shared identity-and-credit foundation.</h3><p><strong>A. D. Togonu-Bickersteth</strong></p><p><em>This is essay 3 of four in the series “Capability for the AI Decade”, drawn from the working paper</em> Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade <em>(2026, </em><a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade"><em>Pulse Public Library</em></a><em>). Read in any order — each essay stands alone — or read in sequence: Essay 1 (</em><a href="https://medium.com/human-at-work/will-the-missing-dimension-in-capability-development-49ca3831973d"><em>Will</em></a><em>), Essay 2 (</em><a href="https://medium.com/human-at-work/schema-as-pedagogy-why-ai-generated-learning-drifts-and-how-to-stop-it-16de078757fd"><em>Schema-as-Pedagogy</em></a><em>), Essay 3 (Three Platforms, One Substrate), Essay 4 (Jurisdictional Fidelity).</em></p><h3>The three problems and why one architecture cannot solve them all</h3><p>Imagine a graduate of a partner university in an emerging-market economy.</p><p>She enters a four-year undergraduate programme in finance at age eighteen. By her final year, she is not the same person who started — not just in what she knows, which is the easy thing to say, but in how she perceives the work she is being prepared for, what she pays attention to in a financial system, and what she would say if asked to advise an institution navigating its first encounter with a regulator. The university’s job has been to develop her into a specific kind of practitioner over four years. The transcript should reflect that becoming. In most institutions, it does not.</p><p>She graduates. Her sponsoring employer enrols her, six months later, into a vocational development programme — a six-month leadership formation cycle, or a regulatory-skills bootcamp, or a transformation programme tied to a strategic priority of the firm. This is not university. The learning problem is different: not the formation of a graduate over four years but the shifting of a working professional’s behaviour in their actual role over six months. The architecture has to be different.</p><p>Three years on, she is a mid-career practitioner. The tools she trained on are obsolescing fast — the analytical platform her firm bought in 2024 is being replaced by something AI-native that does the work in seconds. She joins her firm’s role-mastery cohort: a year-long programme that does not teach her the new tool but teaches her how to make sense of <em>any</em> tool in this category, and the next, and the one after. The learning problem is different again: not behaviour-shift but role mastery in a tool-volatile decade.</p><p>Most learning systems treat these as three different people taking three different courses. They are not. They are one practitioner across the arc of a career, and her becoming is what capability infrastructure has to carry.</p><p>The problem with treating them as three different courses is structural. <em>Designing for Professional Excellence</em> names this dynamic directly: behavioural transformation (vocational and professional) and skill mastery (role-mastery) are two distinct learning problems, with different architectures, and conflating them — the dominant pattern in conventional learning-management infrastructure — neutralises both. The university context adds a third complication: degree-grade higher education has to satisfy regulator and accreditation demands at jurisdictional scale, on top of the developmental architecture itself. Three problems, three architectures.</p><p>But three architectures on three unrelated platforms produce fragmentation. The practitioner moves across courses without continuity. The evidence of her becoming sits in separate, incompatible silos. The architecture that should be carrying her does not exist.</p><p>What follows is an account of three platforms, designed for the three problems separately, sitting on a single substrate that connects them — and what that combination makes possible.</p><h3>EDP Professional— behavioural transformation in vocational and professional development</h3><p>EDP Pro — our vocational and professional development platform — addresses the first of the three learning problems. The learning task is behavioural transformation: shifting how a working professional perceives, interprets, and engages with the work in front of them.</p><p>The architecture is a Learning Brief designed using the CARVE framework (Capabilities, Activities, Results, Value, Evaluation) and generated into a complete programme through a four-phase developmental cycle: <em>Discover, Develop, Deliver, Share</em>. Each phase has a distinct developmental function. <em>Discover</em> surfaces the assumptions and folk models the practitioner brings to the work. <em>Develop</em> introduces the conceptual models and belief reframings that the transformation requires. <em>Deliver</em> moves the learner into practice in real contexts — where the new frame is enacted on actual work, not articulated on the platform. <em>Share</em> reinforces the change through peer accountability and structured reflection.</p><p>Module submission is gated on an Engagement Quality Score that fails if the learner has not engaged across all four phases at threshold quality. The pedagogical philosophy — <em>completion does not equal capability</em> — is enforced as a 400-error rather than asserted as a slogan. Practitioners cannot complete a module by clicking through it. They have to enact, capture, and bring back the enactment for peer review.</p><p>A deliberate-practice loop sits at the heart of every module: <em>Frame, Capture, Reflect, Rehearse, Recommit</em>. Practice threads are bound to specific result-and-activity pairs from the originating CARVE brief, ensuring that practice is always tied back to a CARVE-defined Result and Activity rather than allowed to drift into generic skills-building. The Capture step in particular brings instances of the practitioner’s actual work into the loop — the practitioner records what they did, in their own words and with their own evidence, and the loop is built around that.</p><p>The brief library — more than fifty briefs in production — covers a six-tier leadership ladder, a six-level professional certificate ladder, sector verticals across financial services, AI-first roles (investment analyst, full-stack developer, sales, customer service), and transformation programmes (mental wellness, professional sustainability, life solutions). Each brief is structurally identical at the schema level. Each is substantively distinct in what its CARVE structure encodes.</p><p>The platform’s structural commitment is that behavioural transformation is the learning problem, the four-phase cycle is the developmental architecture, peer-witnessed enactment is the evidence, and the brief library is the corpus that scales the architecture across roles and contexts.</p><h3>EDP Academic — regulator-anchored degree-grade higher education</h3><p>EDP Academic addresses the institutional-grade variant of the same family of learning problems. The university programme has to satisfy regulator and accreditation demands while operating against a developmental architecture that produces actual transformation — not just regulator-approved content delivery.</p><p>The architecture extends the EDP pattern with three additional structural commitments.</p><p>First, <em>the regulator is at the top of the data model</em>. A four-layer schema places the regulator body above everything else: <em>regulator body → regulator programme variant → partner institution → programme → course source record</em>. Every course source record carries a SHA-256 offset hash of the 200-character anchor window in the regulator document. When the regulator releases a new version of the curriculum, the platform can detect “same content, relocated” versus “content changed” with full audit transparency. Source-pointer integrity is not an after-the-fact compliance feature; it is the structural property of the data model. (This is the structural integrity that the fourth essay in this series returns to as the foundation of jurisdictional fidelity.)</p><p>Second, <em>the activity-handle abstraction makes content partner-portable</em>. Content emits handles like {{handle:tutorial_delivery_activity}} rather than partner-specific role names. The activity-handle linter rejects content that names partner-specific roles directly. This means one brief can deploy across multiple partner institutions through configuration alone, not re-authoring. A degree programme designed for the labour structure of one university can be deployed at another whose role labels are different, without any of the substantive content changing.</p><p>Third, <em>the folk-model displacement audit gates run acceptance</em>. A separate model pass evaluates, on every generation, whether the worldview the learner enters with — the common-sense framing of finance, or leadership, or compliance that everyone arrives with — is actually being replaced by the discipline-specific framework. This is the structural answer to the deep failure of regulator-approved curricula that pass accreditation review and still leave graduates with their pre-university folk models intact.</p><p>The corpus already produced demonstrates the architecture at scale: 22 regulator bodies anchored, 26 regulator programme variants, 168 course source records, 35 modules, 8 briefs, 131 production generation runs. Pilot partner cohorts are freshly active.</p><p>The platform’s structural commitment is that degree-grade higher education has to satisfy three commitments simultaneously: regulator authority preserved verbatim, partner-portability built through abstraction, and folk-model displacement gated at the audit layer. None of the three is bolted on. All three are structural to the architecture.</p><h3>ESMP — role mastery for an AI-disrupted decade</h3><p>ESMP — our role-mastery platform — addresses the third learning problem: skill mastery in a decade where the specific tools the practitioner trains on will be obsolete within years.</p><p>The architecture is a six-layer pedagogical taxonomy. At the foundation (L0), the role’s core real-world problem and the category of artefacts that solve it anchor all learning — every cycle on the platform begins from a concrete artefact in the real domain rather than from abstract content. Above L0 sit <em>themes</em> (L1, broad strategic categories), <em>domains</em> (L2, mutually-exclusive knowledge areas), <em>patterns</em> (L3, concrete repeatable techniques), and <em>tools</em> (L4, interchangeable executors). A universal <em>Milestone ladder</em> from M1 (<em>Assisted Executor</em>) to M5 (<em>Strategic Leader</em>) cuts across these layers, describing capability stance independently of the specific tools the practitioner uses.</p><p>The thesis the platform encodes is the AI-decade thesis: <em>the architecture is durable; the tools are interchangeable</em>.</p><p>Two further constructs operate within this taxonomy. <em>Anchors</em> are complete real-world solutions to a role’s core problem at a given milestone — finished artefacts the learner deconstructs, each carrying an instantiation recipe with primary tools, alternative tools, and step-by-step practice scaffolds. <em>Lenses</em> are per-cycle deconstruction frames: given a learning cycle, a focus pattern, and an anchor, a Lens makes the pattern visible within the artefact without requiring a new artefact. One anchor can support many cycles, each lensed through a different pattern.</p><p>The corpus spans nine roles, 44 themes, 77 domains, 305 patterns, 363 tools, and 190 anchors with 2,271 anchor-pattern emphasis mappings. Generation is multi-stage with auto-resilience gap-fill: when initial pattern coverage falls short, the system re-prompts with pattern-type-aware design guidance and runs parallel gap-fill recipes. The architecture-generation function actively auto-repairs LLM output to schema invariants — every pattern has at least two tools, every milestone matrix is complete, every type and level enum is normalised.</p><p>The structural commitment is that role mastery in a tool-volatile decade is not transferred by another course in another tool. It is developed by <em>deconstructing real-world artefacts through different patterns, across the Milestone ladder, on a platform whose architecture is durable enough to survive the next tool transition and the one after</em>.</p><h3>One substrate carrying becoming</h3><p>Three platforms, three architectures, three learning problems. What makes them <em>capability infrastructure</em> rather than three unrelated systems is the shared substrate they sit on.</p><p>Return to the practitioner from the opening.</p><p>She enters EDP Academic at eighteen as a four-year undergraduate. Her record on the platform — her enacted work, the partner faculty’s responses, the folk-model displacement audit results, the regulator-anchored evidence of what she has learned across course source records — accumulates across four years on a single identity, with a single credit substrate behind every AI-generated piece of feedback she received, with a unified audit log of every AI interaction the platform made on her behalf. She graduates with not just a degree but a developmental record.</p><p>She moves into EDP Pro at twenty-three. The same identity carries her there. The credits her employer purchased for her cohort feed the same wallet that buffered her four years on EDP Academic. The AI coach that diagnoses where her growth edge sits across Knowledge, Skill, and Will has access — within the boundaries her data governance settings permit — to the developmental record she brings with her from her undergraduate years. She is not starting from zero with a new system. She is continuing a developmental arc.</p><p>Three years on, she enters ESMP. Again the same identity. The artefacts she deconstructs in role-mastery cycles are evaluated by a mentor who is, on the platform, a senior practitioner two milestones above her. The artefacts she produces during the cycle accumulate in a portfolio that is <em>hers</em>, not a separate cohort credential. By her fifth year as a mid-career professional, the portfolio represents her becoming across all three platforms — degree, vocational, role-mastery — as a single arc.</p><p>This is what the shared identity-and-credit substrate makes possible. It is not a feature. It is the structural condition without which three architectures, however well-designed each one is, fragment into three separate developmental episodes. With it, they constitute <em>infrastructure</em>.</p><p>The architectural insight is straightforward: degree, professional development, and role mastery are three different learning problems that the same practitioner moves across in a continuous developmental arc. The platform infrastructure either carries that arc or it does not. Most institutional learning technology today does not. The three platforms described here, on the substrate they share, do.</p><h3>Where this essay sits</h3><p>This essay is the third of four drawn from the working paper <em>Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade</em> (2026), published in the <a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade">Pulse Public Library</a>. The paper develops the three integrated platforms at greater length as a Design Science contribution — three artefacts instantiating the schema-as-pedagogy pattern on a shared substrate — alongside three other contributions: Will as a developmental dimension (Essay 1), schema-as-pedagogy as a Design Science contribution (Essay 2), and jurisdictional fidelity as a structural property (the subject of Essay 4).</p><p>The previous essay — <em>Schema-as-Pedagogy: Why AI-Generated Learning Drifts, and How to Stop It</em> — argued that pedagogical theory has to be encoded into the data, validators, prompt scaffolds, and tool-use layers of the platform. This essay has shown what schema-as-pedagogy looks like operationalised across three platforms that address three distinct learning problems on a shared identity-and-credit substrate.</p><p>The next essay — <em>Jurisdictional Fidelity: What Legitimate AI-Assisted Education Looks Like Outside the West</em> — addresses the policy-and-legitimacy question that all three platforms have to answer to operate at population scale in non-Western contexts: what makes the infrastructure legitimate in a country whose regulatory authority, cultural register, and institutional norms differ from the West’s, and why decorative localisation is structurally the same failure mode as decorative pedagogy.</p><p>For readers who want to go further, the relevant prior work includes the <a href="https://medium.com/human-at-work/practitioner-intelligence-expertise-reflexive-competence-and-judgment-in-the-age-of-ai-49723fc5d601">Practitioner Intelligence paper</a> (Medium, 2026), which develops the Practitioner Intelligence construct that the three platforms are built to develop; <a href="https://medium.com/human-at-work/epistemic-contact-meaning-judgment-and-the-structural-consequences-of-organisational-scale-a95bc8bcde70">Epistemic Contact</a> (Medium, 2026), which addresses the organisational-scale conditions under which the three architectures are necessary; and the <a href="https://pulse.euchreston.com/library/browse/carve-stop-doing-more-start-achieving-what-matters">CARVE eBook</a> (Pulse, 2026), which is the framework underneath EDP Pro and EDP Academic.</p><p>One substrate, three platforms, three learning problems addressed at population scale. The architecture is what makes a single practitioner’s becoming visible across the arc of a career rather than fragmented across unrelated courses. That arc is what capability infrastructure has to carry.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=47c1c65cf680" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/three-platforms-one-substrate-what-population-scale-capability-infrastructure-looks-like-47c1c65cf680">Three Platforms, One Substrate: What Population-Scale Capability Infrastructure Looks Like</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Schema as Pedagogy: Why AI-Generated Learning Drifts, and How to Stop It]]></title>
            <link>https://medium.com/human-at-work/schema-as-pedagogy-why-ai-generated-learning-drifts-and-how-to-stop-it-16de078757fd?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/16de078757fd</guid>
            <category><![CDATA[professional-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[learning-design]]></category>
            <category><![CDATA[learning-technologies]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Fri, 15 May 2026 07:44:14 GMT</pubDate>
            <atom:updated>2026-05-15T07:44:14.322Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JDrB4dOPhJIi05-Q8VB2rA.png" /></figure><p><em>When the framework is described in marketing material but not enforced at the architecture level, AI-generated content drifts toward generic — and the framework becomes decoration</em></p><p><em>This is essay 2 of four in the series “Capability for the AI Decade”, drawn from the working paper</em> Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade <em>(2026, </em><a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade"><em>Pulse Public Library</em></a><em>). Read in any order — each essay stands alone — or read in sequence: Essay 1 (Will), Essay 2 (Schema-as-Pedagogy), Essay 3 (Three Platforms, One Substrate), Essay 4 (Jurisdictional Fidelity).</em></p><h3>The Drift Problem</h3><p>If you have evaluated an AI-assisted learning platform recently — for institutional procurement, for board oversight, or simply as a candidate for your own development — you will recognise the pattern.</p><p>The vendor demonstration is impressive. There is a well-known capability framework named on the product page — Bloom’s Taxonomy, the 70–20–10 model, a proprietary “transformation methodology”, or some adaptation of an open educational framework. The AI tutor responds quickly and competently. The content reads well. The framework appears in the language the system uses to describe itself: “Each module is grounded in our pedagogical framework.” “Our AI coach personalises learning to your developmental needs.” “We blend Bloom’s Taxonomy with experiential learning.”</p><p>You sign the procurement papers. The cohort starts. Six weeks in, you start reading the actual content the platform is generating for your learners. And you notice something: the framework is not really there.</p><p>The content reads well. The format is right. The headings name the framework. But when you look closely, the substantive structural commitments of the framework — the things that distinguish it from any other coherent piece of writing about leadership or sales or programme management — are missing. The AI produced fluent content; it produced content shaped by the surface conventions of the genre; it produced content that <em>names</em> the framework. It did not produce content that <em>carried</em> the framework.</p><p>This is what I call the drift problem. It is the dominant failure mode of AI-paste-on educational platforms today, and it is structural rather than accidental. Generative AI systems are optimised for fluency over fidelity. When asked to produce educational content, they will produce content that reads well and conforms to surface conventions of the genre. They will not, of their own accord, conform to the structural commitments of any particular pedagogical framework. The framework, if it is not explicitly enforced at the architecture level, becomes decoration.</p><p>I will argue in this essay that the only structural answer to this problem is what I call <em>schema-as-pedagogy</em> — a design pattern in which pedagogical theory is encoded into the data, validators, prompt scaffolds, and tool-use schemas of the platform itself, such that the framework cannot drift because the system normalises every piece of generated content to its own invariants.</p><p>This is not a problem you can solve by writing better prompts. It is not a problem you can solve by hiring more pedagogists to review the output. It is a problem you can only solve by encoding the framework at the architecture level — and once you do, the question of which platforms should be procured for institutional use becomes much sharper.</p><h3>Why this Problem Cannot be Solved by Better Prompts Alone</h3><p>The first response of most teams who recognise the drift problem is to try to solve it with better prompts. If the AI produced generic content, write a more specific prompt. If the AI did not apply the framework, tell it to apply the framework. Better prompts, the thinking goes, will produce better content.</p><p>This response is wrong, but it is wrong in instructive ways.</p><p>The first wrongness is that prompts have no enforcement. A prompt is a request, not a constraint. The LLM may follow the request, may follow part of it, may interpret it in a way the prompt-author did not intend. If the prompt says “apply the K/S/W framework”, the AI will produce content that mentions K/S/W; whether that content actually applies the K/S/W framework — whether each capability is decomposed across Knowledge, Skill, and Will, with the structural sub-properties of each — is something the prompt cannot guarantee.</p><p>The second wrongness is that drift compounds at scale. A single generation can be reviewed by a human pedagogist; a thousand generations cannot. If your platform is generating content for fifty courses across twenty programmes for three partner institutions, the drift in any given module is not catastrophic on its own. The drift across the whole corpus is what destroys the framework. By the time the drift is visible at the corpus level, the platform has produced thousands of pieces of generic content whose marketing copy claims a framework they do not embody.</p><p>The third wrongness is that the LLM’s underlying optimisation is not for fidelity. The LLM is trained to produce fluent text that satisfies surface conventions. The framework is a structural commitment. Fluent satisfaction of surface conventions and structural commitment to a specific framework are not the same thing. Asking the LLM to produce both, through the prompt alone, is asking it to satisfy a requirement that conflicts with its underlying training. It will satisfy the easier of the two — fluency — and let the harder one drift.</p><p>The architectural alternative starts from a different premise. It assumes the LLM cannot be trusted to enforce the framework. It assumes the LLM is, and will remain, optimised for fluency. And it builds the platform around that assumption — encoding the framework at the data, validator, prompt-scaffold, and tool-use layers, so that what the LLM produces cannot leave the system unless it conforms to the framework’s structural commitments. The system does not ask the model to follow the framework. The system normalises the model’s output to its own invariants.</p><p>This is what I mean by <em>schema-as-pedagogy</em>.</p><h3>The Four Layers of Schema-as-Pedagogy</h3><p>In our work at Euchreston — across the EDP Pro, EDP Academic, and ESMP platforms — we have come to think of schema-as-pedagogy as a four-layer architectural commitment. Each layer addresses a different way the framework could fail to be enforced. Together they constitute the structural alternative to AI-paste-on platforms.</p><h4><strong>Layer One: Data.</strong></h4><p>The schema of the platform’s data structures — the database tables, the JSON shapes, the document types — encodes the framework. Required fields are required because the framework requires them. Sub-structures within fields are present because the framework’s structural commitments are themselves structured.</p><p><strong>In EDP Pro </strong>— our vocational and professional development platform — every Learning Brief is a CARVE-structured document: it must have Capabilities (with the K/S/W decomposition), Activities (with situational judgment elements), Results (with three gap types and four levels of completion-to-effectiveness), Value (across four levels), and Evaluation (with breakdown, deliberate practice, reflective practice, and reflexive competence). A document cannot be saved as a Learning Brief without these fields populated. The framework is not described in the brief; the framework is the schema the brief inhabits.</p><p><strong>In EDP Academic</strong> — our degree-grade higher education authoring platform — the four-layer course-source-record schema places the regulator at the top of the data model: <em>regulator body → regulator programme variant → partner institution → programme → course source record</em>. Every course source record carries a SHA-256 offset hash of the 200-character anchor window in the regulator document, ensuring source-pointer integrity that survives regulator-version changes. Jurisdictional context is not decoration. It is the data model.</p><p><strong>In ESMP</strong> — our role-mastery platform — the six-layer pedagogical taxonomy (themes, domains, patterns, tools, milestones M1–M5, with anchors and lenses operating across the taxonomy) is enforced at the data layer. Every pattern must have at least two tools. Every milestone matrix must be complete. Every type and level enum is normalised at write time.</p><p>The structural commitment is: <em>if a field is missing, the framework is missing, and the data cannot be saved</em>.</p><h4><strong>Layer Two: Validators.</strong></h4><p>Rule-based and LLM-judge validators run on content as it moves through the system. They are gates. Content that does not pass cannot proceed.</p><p>In EDP Academic, the regulator-immutability validator detects and rejects paraphrasing of verbatim regulator text — preserving regulator authority intact. The activity-handle linter detects and rejects content that names partner-specific roles directly; the system requires abstracted activity handles instead, which makes content portable across partner institutions. The folk-model displacement audit, which runs as a separate model pass, evaluates whether the worldview the learner enters with is actually being replaced by the discipline-specific framework — gating run acceptance on this dimension.</p><p>In ESMP, the architecture-generation function actively auto-repairs LLM output to schema invariants — adding missing tools, normalising type enums, completing milestone matrices that the model produced incomplete.</p><p>The structural commitment is: <em>content that violates the framework is detected and rejected before it leaves the validation stage</em>.</p><h4><strong>Layer Three: Prompt Scaffolds.</strong></h4><p>The prompts the AI sees include the framework as structural constraint, not as suggestion. The AI is not asked to consider the framework; the AI is constrained, by the prompt structure itself, to operate within the framework.</p><p>When EDP Pro generates a Learning Brief, the prompt scaffold requires the model to produce content for every CARVE field. It does not tell the model “please consider including the K/S/W decomposition”; it gives the model a structured form in which the K/S/W slots are present and must be populated. The model can produce poor content for those slots — that is what the validator layer catches — but it cannot omit them.</p><p>In ESMP, when initial pattern coverage falls short, the system re-prompts with pattern-type-aware design guidance and runs parallel gap-fill recipes. The prompt itself adapts to the gap.</p><p>The structural commitment is: <em>the framework is in the prompt, so the AI cannot proceed without applying it</em>.</p><h4><strong>Layer Four: Tool-Use.</strong></h4><p>The AI can only do what the tools permit it to do. The tools enforce the schema at the action level, not just at the content level.</p><p>When the platform calls the LLM, it provides a fixed set of function-calling tools — and each tool enforces a slice of the schema. There is no “save freeform content” tool. There is “save a CARVE-structured Learning Brief”, or “save a course source record with regulator anchor”, or “save a pattern with at least two tools”. The AI cannot bypass the schema by saving content in a free-form way, because the tool to do so does not exist.</p><p>The LLM-judge validators that run as separate model passes are themselves tool-bound — they are constrained to specific evaluation tasks, with structured output schemas, and their output is itself normalised to the platform’s invariants.</p><p>The structural commitment is: <em>the AI’s action surface is the schema; the AI cannot act outside it</em>.</p><p>Together, these four layers normalise every piece of generated content to the framework’s invariants. The system does not trust the model. The system enforces the pedagogy.</p><h3>What Changes for Procurement and Institutional Partners</h3><p>If schema-as-pedagogy is the architectural alternative to drift, then a clean test follows for institutional buyers evaluating AI-assisted learning platforms.</p><p>The test is not “does the vendor name a pedagogical framework?” Almost every vendor names one. The test is also not “does the AI produce content that mentions the framework?” Almost every AI does. The test is: <em>show me where the framework is enforced</em>.</p><p>Specifically, four questions a procurement committee can ask:</p><p>1. <strong>At the data layer.</strong> Is the framework’s structural commitment enforced as required fields in the platform’s data model? If a piece of content is missing the fields the framework requires, can it be saved? If yes, the framework is decoration.</p><p>2. <strong>At the validator layer.</strong> What rules detect content that violates the framework? Are those rules rule-based, LLM-judge, or both? Show me a rejection trace from a real generation run.</p><p>3. <strong>At the prompt-scaffold layer.</strong> What does the AI actually see when it is asked to generate content? Is the framework in the prompt as a structural form, or is it a free-text instruction the AI can interpret?</p><p>4. <strong>At the tool-use layer.</strong> What tools does the AI have access to, and what do those tools enforce? Can the AI save content outside the schema?</p><p>A vendor whose platform genuinely encodes the framework at the architecture level will be able to answer all four questions concretely, with reference to actual code and database schemas. A vendor whose framework is decorative will answer with marketing language. The difference is visible in five minutes.</p><p>This test matters more in non-Western contexts than in Western ones, for reasons I will develop in the fourth essay of this series. Briefly: when the institutional context places jurisdictional and cultural constraints on what the framework should encode — Nigerian regulator authority, British English spelling, naira-and-kobo currency, Singapore’s PDPA versus the EU’s GDPR — the only way to enforce those constraints across a large generated corpus is to encode them in the architecture. Decorative localisation is the same failure mode as decorative pedagogy. The structural answer is the same.</p><p>For institutional partners — universities, foundations, multilaterals, government bodies — the implication is that the procurement decision is not principally a feature decision. It is an architectural decision. The platform you choose locks in the structural commitments your population’s learning will be built on. If those structural commitments are decorative, you will get a fluent corpus that satisfies surface conventions and produces nothing in cohort. If they are encoded, you will get a corpus that carries what the framework promises.</p><p>This shift — from feature-procurement to architecture-procurement — is one of the most consequential changes in institutional learning technology in this decade. The next essay in this series shows what the architecture looks like when it is built around three different learning problems on a shared substrate, and how the practitioner moves through it across the arc of a career.</p><h3>Where this Essay Sits</h3><p>This essay is the second of four drawn from the working paper <em>Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade</em> (2026), published in the <a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade">Pulse Public Library</a>. The paper develops schema-as-pedagogy as a Design Science Research contribution at greater length, alongside three other contributions: Will as a developmental dimension (the subject of the previous essay), the three integrated platforms presented as a working operational realisation, and jurisdictional fidelity as a structural property.</p><p>The previous essay — <a href="https://medium.com/human-at-work/will-the-missing-dimension-in-capability-development-49ca3831973d"><em>Will: The Missing Dimension in Capability Development</em> </a>— argued that <em>Will</em> is the developmental dimension that conventional competency models systematically neglect, and that taking Will seriously requires a developmental architecture rather than a measurement instrument. This essay has argued that the developmental architecture cannot remain a structural commitment in design documents — it has to be enforced at the data, validator, prompt-scaffold, and tool-use layers of the platform itself. Without that enforcement, AI-generated content drifts toward generic, and the framework becomes decoration.</p><p>The next essay — <em>Three Platforms, One Substrate</em> — turns to the operational realisation. It tells the story of how three integrated platforms, each addressing a distinct learning problem, sit on a shared identity-and-credit substrate, and how a single practitioner’s becoming is carried across all three over the arc of a career.</p><p>For readers who want to go further, the relevant prior work includes <a href="https://www.linkedin.com/pulse/designing-professional-excellence-euchreston-kcvac/">Designing for Professional Excellence</a> (LinkedIn, 2026), which sets out the structural distinction between behavioural transformation and skill mastery that the EDP and ESMP architectures address; and the <a href="https://pulse.euchreston.com/library/browse/carve-stop-doing-more-start-achieving-what-matters">CARVE eBook</a>, which develops the framework whose architectural enforcement this essay describes.</p><p>Pedagogical theory is not what you say in the marketing material. It is what the architecture enforces. And until institutional buyers ask vendors to show where the framework is encoded — at the data layer, at the validator layer, at the prompt scaffold, at the tools — the AI-assisted learning market will continue to ship platforms that look modern in a demo and produce nothing in cohort.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=16de078757fd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/schema-as-pedagogy-why-ai-generated-learning-drifts-and-how-to-stop-it-16de078757fd">Schema as Pedagogy: Why AI-Generated Learning Drifts, and How to Stop It</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Will: The Missing Dimension in Capability Development]]></title>
            <link>https://medium.com/human-at-work/will-the-missing-dimension-in-capability-development-49ca3831973d?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/49ca3831973d</guid>
            <category><![CDATA[leadership-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[capability-development]]></category>
            <category><![CDATA[learning-and-development]]></category>
            <category><![CDATA[higher-education]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Fri, 08 May 2026 06:15:42 GMT</pubDate>
            <atom:updated>2026-05-08T06:15:42.434Z</atom:updated>
            <content:encoded><![CDATA[<h4>We have Knowledge. We have Skill. We are missing the dimension of intentional choice — and we have to design for it, not measure it.</h4><p><strong>A. D. Togonu-Bickersteth</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NWbd6YpGAX2rKLwgc9eELg.png" /></figure><p><em>This is essay 1 of four in the series “Capability for the AI Decade”, drawn from the working paper</em> Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade <em>(2026, </em><a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade"><em>Pulse Public Library</em></a><em>). Read in any order — each essay stands alone — or read in sequence: Essay 1 (Will), Essay 2 (Schema-as-Pedagogy), Essay 3 (Three Platforms, One Substrate), Essay 4 (Jurisdictional Fidelity).</em></p><p>Most competency frameworks I have ever read measure two things: what the practitioner knows and what the practitioner can do. Both are important. Neither tells me whether the practitioner has chosen to become the practitioner the work needs them to be — and <em>that</em> is the dimension that matters most: the one we cannot solve by adding a rubric, the one we have to design for.</p><p>If you open any modern competency framework — leadership development matrices, professional certification ladders, graduate scheme assessment criteria, the rubric your university used to grade your final-year capstone — you will find that it measures along two axes. The first is <em>Knowledge</em>: the propositional content the practitioner can reproduce, recognise, or apply. The second is <em>Skill</em>: the operational competence the practitioner can demonstrate when given a task. The framework will likely structure these as a hierarchy — basic to advanced, novice to expert, levels one through five — and provide rubrics for assessing each.</p><p>What this two-dimension model assumes is that if you measure Knowledge and Skill accurately enough, you have measured capability. You haven’t. You have measured what the practitioner can <em>do</em> under instruction. You have not measured whether the practitioner <em>will</em> — when nobody is watching, when the task is messy and the conditions are imperfect, when the right thing to do is not the easy thing — choose to act in a way that grows them into the practitioner the work needs them to be.</p><p>That choice is the dimension I want to name <em>Will</em>. And the fact that conventional competency frameworks systematically leave it out is not an oversight. It is structural. Will is hard to instrument. It is hard to grade. It is hard to compare across cohorts. So competency frameworks have tended either to ignore it entirely or to wave at it through proxy constructs — <em>motivation</em>, <em>engagement</em>, <em>character</em>, <em>grit</em> — and then move on.</p><p>This essay argues that Will is the developmental dimension that matters most for the AI decade. It is the dimension that distinguishes the practitioner who completes the programme from the practitioner who has actually become someone different by the end of it. And taking Will seriously requires us to stop trying to measure it and start <em>designing for</em> it.</p><h3>Will is not motivation, character, or grit</h3><p>Before I make the design argument, I have to push back on the proxies. When practitioners hear “Will” they reach reflexively for adjacent constructs — motivation, character, grit, drive, attitude, mindset. These are all real. None of them is what I mean by Will.</p><p><em>Motivation</em> is a state. It varies day to day, hour to hour. The practitioner who is highly motivated on Monday morning may be exhausted by Thursday evening. Motivation is responsive to mood, context, sleep, the last conversation, the next deadline. It is real, and it matters, but it is not the dimension on which capability development should rest. If you build a developmental architecture that depends on the practitioner being highly motivated when they encounter the work, you have built fragility into the architecture. Motivation will fail you. The practitioner you are trying to develop has the rest of their life happening around them.</p><p><em>Character</em> is a summary. We say someone has “good character” when we mean the long-term aggregate of their choices has produced a recognisable pattern. But character is what we observe after the fact. It is not the dimension of choice that produces the pattern. By the time you are evaluating someone’s character, the developmental work is done — for better or worse.</p><p><em>Grit</em> is closer. The disposition to persist toward a long-term goal in the face of obstacles is genuinely close to what I am pointing at. But grit, as the construct has been operationalised, treats persistence as a trait — something the practitioner has more or less of, like height or extraversion. That is not how Will operates. Will is not a trait. It is a <em>choice</em>, made and remade, in the conditions of practice.</p><p>What I mean by Will is what Aristotle pointed at with <em>prohairesis</em> — the deliberate choice that constitutes character through repeated exercise. The classical sources are clear that the choice is what makes the character, not the other way around. You do not develop Will by being told to “have more grit”. You develop Will by being placed, again and again, in conditions where the choice has to be made — and by being held to the choice when you make it.</p><p>This is the structural insight that the proxy constructs miss. Will is not a quality the practitioner brings to the work. It is a dimension of the work itself. It is exercised, every time, by the practitioner who chooses — when the lecture material would have been enough, when the certificate is already in hand, when no one is watching — to take the next step in becoming the practitioner the work needs them to be.</p><p>And once you see Will this way, the implications for capability development are direct. You cannot install Will in someone the way you install Knowledge. You cannot drill Will into someone the way you drill Skill. What you can do is build the conditions in which Will is exercised, capture the choices the practitioner makes when they are exercised, and treat the pattern of those choices as the developmental record. That is what I mean by <em>designing for</em> Will.</p><h3>Three layers of designing for Will</h3><p>So how do you design for a dimension that, by definition, cannot be drilled or instrumented? The answer is to architect the conditions in which the choice has to be made, observed, and accumulated. In our work at Euchreston — across the EDP and ESMP platforms — we have come to think of this as a three-layer architecture. Each layer addresses one structural requirement of designing for Will. Together they constitute the developmental architecture.</p><h4><strong>Layer one: designed enactment.</strong></h4><p>The practitioner has to enact the transformation in real conditions, not write about it.</p><p>In EDP Pro — our vocational and professional development platform — every learning module follows a four-phase developmental cycle: <em>Discover</em>, <em>Develop</em>, <em>Deliver</em>, <em>Share</em>. The first two phases are conceptual: surfacing assumptions in <em>Discover</em>, building new frames in <em>Develop</em>. But the developmental work happens in <em>Deliver</em> — where the practitioner takes the new frame into their actual work and acts on it. The deliberate-practice loop that sits inside Deliver — Frame, Capture, Reflect, Rehearse, Recommit — brings real-world instances of the practitioner’s practice into the loop. The capture is not hypothetical. The reflection is not abstract. The rehearsal is anchored to a specific result-and-activity pair from the practitioner’s actual work.</p><p>In ESMP — our role-mastery platform — every learning cycle is anchored to a real-world artefact: a deliverable that practitioners in this role actually produce in the field. Each cycle deconstructs an existing artefact through a chosen pattern (a <em>Lens</em>) and then asks the practitioner to produce or revise an artefact themselves, applying the pattern. There is no abstract problem set. There is the artefact, the pattern, and the practitioner’s enactment.</p><p>The structural commitment is that the platform never substitutes a representation of the work for the work itself. You cannot complete the module by reading about the deliverable. You complete it by producing one — or by enacting the changed practice in the world where it counts.</p><h4><strong>Layer two: witnessed observation.</strong></h4><p>The practitioner’s enacted work is observed by real people who can respond.</p><p>In EDP Pro the <em>Share</em> phase is where this happens. The practitioner brings their enacted practice to peers in the cohort — a presentation, a portfolio of practice instances, a structured reflection on what shifted. Peers observe and respond. The platform’s role is not to evaluate the work; the platform’s role is to surface the work to people who can.</p><p>In ESMP the witness configuration is different. The practitioner builds a portfolio of artefacts over multiple cycles and a <em>mentor</em> — typically a practitioner already operating at a higher Milestone in the same role — reviews the portfolio in structured dialogue. The mentor is not grading. The mentor is observing the developmental pattern across the portfolio and naming what they see. The practitioner is asked to respond.</p><p>In EDP Academic — our degree-grade higher education platform — the witness configuration is the partner institution’s faculty. The practitioner’s enacted work in their programme is reviewed by faculty who themselves are operating in the field of practice the programme is preparing the student for. The audit substrate behind EDP Academic ensures that what is reviewed is the practitioner’s actual work, not a proxy for it.</p><p>The structural commitment is that the witness role is held by a real person — peer, mentor, or faculty member — not by a rubric. The platform creates the social-accountability structure. People do the witnessing.</p><h4><strong>Layer three: accumulated portfolio.</strong></h4><p>What builds over time is the developmental record.</p><p>A practitioner working through EDP Pro accumulates, across modules, a record of practice instances they captured, reflections they made, peers’ responses to those reflections, and recommitments they articulated. That accumulated record is the developmental evidence — not the completion certificate, not the engagement quality score, not the sequence of module entries.</p><p>A practitioner working through ESMP accumulates, across cycles, a portfolio of artefacts produced and revised, with the mentor’s responses recorded against each. That portfolio is what represents the practitioner’s becoming over the arc of their development on the platform.</p><p>In EDP Academic, the accumulated portfolio is the practitioner’s transcript of practice — the record of what they did across the programme, reviewed by faculty, anchored to the regulator-approved outcomes the programme is built around but exceeding what those outcomes alone would capture.</p><p>The structural commitment is that the developmental record is not the platform’s grade-book. It is the practitioner’s enacted body of work, reviewed by the people who watched it be made, and accumulated over time as the answer to the only question that matters: did this practitioner become someone different by the end of the programme?</p><h3>What changes if developmental architectures were designed for Will</h3><p>If you take Will seriously as a developmental dimension and you build for it through the three-layer architecture, several things change.</p><p>For <em>university curriculum design</em>, the implication is that a degree programme is not a content list to be delivered. It is a developmental arc to be designed. The question is no longer “what regulator-approved outcomes does the course cover?” — important as that question remains for accreditation. The deeper question is: across the four years of the programme, what is the practitioner becoming, and what is the architecture that ensures that becoming is enacted, witnessed, and accumulated rather than merely tested? Programmes that take this question seriously will look very different from programmes that satisfy themselves with the regulator-approved outcomes alone.</p><p>For <em>professional certification</em>, the implication is more uncomfortable. Most professional certifications today operate on an exam-based proxy. The candidate sits an examination; if they pass, they are certified. The exam is a sample of Knowledge and Skill at a moment in time. It tells you almost nothing about Will. Designing for Will means certifying against evidence of changed practice in the field over time — a portfolio reviewed by qualified witnesses — not against a one-shot exam. This is a harder evidentiary bar. It is also a more honest one. The certifying body that takes this seriously is going to have to accept that certification looks more like ongoing recognition of practice than like a moment of qualification.</p><p>For <em>AI-decade reskilling</em>, the implication is that course catalogues are not the answer. The half-life of any specific tool is collapsing; the practitioner who last year was certified in Tool A is by next year being asked to operate in Tool B. What survives the tool transition is the practitioner’s <em>capability stance</em> — their developed pattern of judgment about what kinds of problems this category of tool is for, and how to make the problem visible enough to the tool to get useful work back. Capability stance is not transferred by another course. It is developed by designed practice loops anchored to the practitioner’s actual work, witnessed by people who are themselves operating at a higher capability stance, accumulated over time. This is not something you can scale through a catalogue. It is something you can scale through architecture.</p><p>For <em>accreditors</em> — the institutions whose recognition makes a credential worth holding — the implication is that the audit shifts. The audit question of “did the learner complete the syllabus?” becomes “did the learner’s practice in the world change in ways the architecture predicted?” That is a harder evidentiary bar but a more meaningful one. And the audit substrate that supports it has to be designed in: source-pointer integrity, witness records, accumulated portfolio, time-stamped responses. The audit cannot be retrofitted. It has to be the structural commitment of the platform itself — which is the subject of the next essay in this series.</p><h3>Where this essay sits</h3><p>This essay is the first of four drawn from the working paper <em>Operationalising Practitioner Intelligence: A Design Science Account of Population-Scale Capability Infrastructure for the AI Decade</em> (2026), published in the <a href="https://pulse.euchreston.com/library/browse/operationalising-practitioner-intelligence-a-design-science-account-of-population-scale-capability-infrastructure-for-the-ai-decade">Pulse Public Library</a>. The paper develops Will as a developmental dimension at greater length, alongside three other contributions: schema-as-pedagogy as a Design Science contribution, three integrated platforms presented as a working operational realisation, and jurisdictional fidelity as a structural property of legitimate AI-assisted educational infrastructure in non-Western contexts.</p><p>The next essay in this series — <em>Schema-as-Pedagogy: Why AI-Generated Learning Drifts, and How to Stop It</em> — addresses the structural-architectural commitment that makes designing for Will (and, more broadly, designing for any pedagogical theory) possible at AI-generated scale. The deeper argument behind both essays is the same: pedagogical theory has to be encoded into the architecture of the system, not declared in the marketing material. This essay has been about <em>what</em> you encode when you take Will seriously. The next is about <em>how</em> the encoding survives at scale.</p><p>For readers who want to go further, the relevant prior work includes the <a href="https://medium.com/human-at-work/practitioner-intelligence-expertise-reflexive-competence-and-judgment-in-the-age-of-ai-49723fc5d601">Practitioner Intelligence paper</a> (Medium, 2026), which develops the construct of <em>reflexive competence</em> that informs the K/S/W decomposition; the <a href="https://medium.com/human-at-work/mental-models-and-belief-systems-20e61506d825">Two-Pronged Approach to Behavior Change</a> (Medium, 2025), which addresses mental models and belief systems as the substrate Will operates on; and the <a href="https://pulse.euchreston.com/library/browse/carve-stop-doing-more-start-achieving-what-matters">CARVE eBook</a>, which develops the framework that informs the platforms named here.</p><p>Will is not soft. It is structural. The question is not how we measure it but how we design for it — and the evidence of whether the design worked sits in the practitioner’s enacted practice in the world, witnessed and responded to by the people they work with, accumulated over time as a portfolio of who they have become.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=49ca3831973d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/will-the-missing-dimension-in-capability-development-49ca3831973d">Will: The Missing Dimension in Capability Development</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Capability in the Age of AI]]></title>
            <link>https://medium.com/human-at-work/capability-in-the-age-of-ai-a045e79e0531?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/a045e79e0531</guid>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[professional-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Tue, 05 May 2026 10:53:39 GMT</pubDate>
            <atom:updated>2026-05-05T10:53:39.530Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>Why Execution Is Being Automated and Judgment Is Becoming the Work</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/580/1*1FZAkezIyTvU0XR7CL6oAA.png" /></figure><p><em>This article is part of my ongoing reflection on how AI is reshaping work, informed not only by observation but by direct, first-hand experience of leaning heavily on AI tools in day-to-day professional activity. As I increasingly rely on these systems, particularly in the context of natural language programming and building real solutions from ideas, what becomes visible is not simply an increase in efficiency, but a more fundamental shift in how capability itself is expressed.</em></p><p>For most professionals, capability has long been defined in a way that assumes value is created through the application of domain expertise. Individuals build knowledge within a field, develop fluency in its methods and frameworks, and demonstrate capability through consistent execution. This model has been effective because work itself was structured around the reliable application of knowledge to known classes of problems.</p><p>What is changing is not the need for outcomes, but the location of effort required to produce them. As AI systems become increasingly capable of handling bounded decision-making within domains, much of what was previously considered expert execution becomes accessible at scale. The implication is not that expertise becomes irrelevant, but that its primary expression shifts.</p><h3>The center of gravity moves from execution to judgment.</h3><p>The distinction is subtle but consequential. Execution is concerned with how to do something within a defined space, while judgment is concerned with what should be done in a specific context where conditions are incomplete, ambiguous, or evolving. As long as work remained largely within bounded domains, execution was a sufficient proxy for capability. As work becomes increasingly shaped by context, that proxy begins to break.</p><p>This has a direct impact on how domains function as a source of professional value. Domain knowledge continues to matter, but it no longer differentiates in the same way when it is expressed primarily through constrained decision spaces. Even when those decisions are complex and historically required deep expertise, their bounded nature makes them amenable to modeling and automation. As a result, the value of domain expertise increasingly depends on how it is applied beyond its formal boundaries rather than within them.</p><p>Real-world problems do not present themselves as well-defined cases that map neatly onto a single domain. They are dynamic, incomplete, and often contradictory, requiring the integration of multiple forms of knowledge and the navigation of competing priorities. In this environment, value is created less by executing within a domain and more by exercising judgment across contexts.</p><h3>This is where the concept of Practitioner Intelligence becomes central.</h3><p><a href="https://medium.com/human-at-work/practitioner-intelligence-expertise-reflexive-competence-and-judgment-in-the-age-of-ai-49723fc5d601">Practitioner Intelligence</a> is the capacity to reliably deliver value in real conditions by integrating knowing, working, and doing. Knowing refers to conceptual understanding, including frameworks, models, and principles drawn from one or more domains. Doing refers to the execution of actions that produce tangible outcomes. Working sits between the two and represents the ability to interpret situations, make sense of incomplete information, and determine how knowledge should be applied in context.</p><p>In traditional models of capability, knowing and doing are emphasized because they are easier to teach, assess, and scale. Working, which is where judgment resides, is often underdeveloped because it requires exposure to real situations where decisions carry consequences.</p><h3>AI changes this balance in two ways.</h3><p><strong>First, </strong>it reduces the relative importance of knowing and doing when they are expressed within bounded domains. AI systems can increasingly provide access to knowledge on demand and can execute or support execution with a level of speed and consistency that is difficult to match.</p><p><strong>Second, </strong>and more importantly, it creates new conditions under which working, or judgment in context, can be developed and expressed more effectively.</p><p>When used deliberately, AI can act as a cognitive amplifier for Practitioner Intelligence. It can surface relevant knowledge across domains, simulate alternative approaches, stress-test assumptions, and extend the range of scenarios that an individual can explore. In doing so, it reduces the cognitive load associated with execution and information retrieval, allowing more attention to be directed toward interpretation, decision-making, and consequence management.</p><p>In this sense, AI does not simply replace elements of capability. It redistributes them and, in doing so, creates the possibility for individuals to operate more consistently at the level of judgment rather than execution.</p><h3>This leads to a clearer distinction in how capability should be understood and developed.</h3><p><strong>Capability as execution</strong> is grounded in learning how to do within a domain. It involves acquiring the ability to perform tasks, apply methods, and follow established processes with accuracy and efficiency. This remains necessary, but it is no longer sufficient as the primary basis for differentiation.</p><p><strong>Capability as judgment</strong> is grounded in learning how to operate within real-world contexts. It involves interpreting situations, evaluating trade-offs, adapting to changing conditions, and making decisions that align with desired outcomes under uncertainty. This form of capability is inherently less structured, more context-dependent, and more difficult to codify, but it is increasingly where value is created.</p><p>The implication is not that individuals should abandon domain expertise, but that they should reframe how it is used. Domain knowledge becomes an input into judgment rather than the primary output of capability. Its value is realized when it is combined with other forms of knowledge and applied appropriately in context.</p><p>For individuals, this requires a shift in how capability development is approached. The focus needs to move beyond accumulating knowledge and refining execution toward building the ability to operate effectively in conditions where problems are not predefined and solutions are not obvious. This involves developing comfort with ambiguity, the ability to integrate multiple perspectives, and the discipline to make decisions in the absence of complete information while remaining accountable for outcomes.</p><p>For organizations, the challenge is more structural. Systems that continue to define and measure capability primarily in terms of execution within domains risk producing individuals who are highly competent in areas that are increasingly susceptible to automation. Developing Practitioner Intelligence requires environments that expose individuals to real problems, encourage cross-domain integration, and recognize the role of context in shaping outcomes.</p><p>In this context, the distinction between capability as execution and capability as judgment becomes a practical lens for understanding how professional value is evolving. Execution does not disappear, but it becomes increasingly supported by AI and less central as a source of differentiation. Judgment, by contrast, becomes the primary locus of value because it is where decisions carry consequence and where context must be interpreted rather than assumed.</p><p>What emerges is a different understanding of capability, one that is less about what an individual can do in isolation and more about how effectively they can integrate knowledge, action, and context to produce meaningful outcomes.</p><p>AI, when approached in this way, is not simply a force that displaces human capability. It becomes a mechanism that enables a more advanced form of it, provided individuals and organizations are willing to shift their focus from execution within domains to judgment within contexts.</p><h3><strong>References and Intellectual Foundations</strong></h3><p>This perspective builds on established work across decision-making, expertise, learning, and organizational theory:</p><p>Herbert A. Simon (1960, 1977), <em>Administrative Behavior</em>; <em>The Sciences of the Artificial: </em>Foundational work on bounded rationality and decision-making as the core of professional activity.</p><p>Gary Klein (1998), <em>Sources of Power: How People Make Decisions: </em>Naturalistic decision-making and the role of experience in real-world judgment.</p><p>Daniel Kahneman (2011), <em>Thinking, Fast and Slow: </em>Dual-process theory and the cognitive mechanisms underlying judgment.</p><p>Donald Schön (1983), <em>The Reflective Practitioner: </em>Professional practice as reflection-in-action within complex, uncertain environments.</p><p>David Kolb (1984), <em>Experiential Learning: </em>Learning as a process grounded in experience, reflection, and application.</p><p>Etienne Wenger (1998), <em>Communities of Practice: </em>Situated learning and the importance of context in capability development.</p><p>Michael Polanyi (1966), <em>The Tacit Dimension: </em>Tacit knowledge and the limits of explicit, codified understanding.</p><p>Hubert and Stuart Dreyfus (1986), <em>Mind Over Machine: </em>Skill acquisition and the progression from rule-based to intuitive expertise.</p><p>Erik Brynjolfsson and Andrew McAfee (2014, 2017), <em>The Second Machine Age</em>; <em>Machine, Platform, Crowd: </em>Economic and organizational implications of digital and AI technologies.</p><p>Richard Rumelt (2011), <em>Good Strategy/Bad Strategy: </em>The role of diagnosis and judgment in effective strategic decision-making.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a045e79e0531" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/capability-in-the-age-of-ai-a045e79e0531">Capability in the Age of AI</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Keeping Your Job in the Age of AI]]></title>
            <link>https://medium.com/human-at-work/keeping-your-job-in-the-age-of-ai-5a20c921b602?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/5a20c921b602</guid>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[careers]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Fri, 01 May 2026 08:05:42 GMT</pubDate>
            <atom:updated>2026-05-01T08:05:42.338Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*88PzgAnA5Ry56B7L0Oh1Ig.png" /></figure><h3>Why the Real Shift Is About Decisions, Not Tasks</h3><p>For much of the past decade, the conversation about AI and the future of work has been framed in relatively simplistic terms. The dominant question has been whether AI will replace jobs, and if so, which roles are most at risk and which might remain relatively protected. This framing is intuitive, but it is also increasingly inadequate for understanding what is actually happening.</p><p>The more important shift is not at the level of jobs, and not even at the level of tasks. It is at the level of decisions.</p><p>AI is not simply automating work; it is progressively absorbing the decisions that sit within work. Once this becomes clear, the question of “keeping your job” takes on a very different meaning. The issue is no longer whether a role disappears in a binary sense, but whether your contribution is anchored in parts of the work that are being absorbed, reshaped, or elevated as decision-making itself is redistributed between humans and machines.</p><p>Every piece of work, regardless of how simple or complex it appears on the surface, is structured by a series of decisions. Some of these decisions are tightly bounded and operational in nature, such as determining how to classify an incoming request, whether a transaction falls within predefined limits, or which template should be applied in a given situation. Others are more expansive, involving judgments about how work should flow, how exceptions should be handled, and how different actors or systems should coordinate over time. At the highest level, there are decisions that define what the work is ultimately trying to achieve, including what should be prioritized, which trade-offs are acceptable, and what constitutes success in a particular context.</p><p>Historically, these decisions have been made by humans not because all of them required deep expertise or judgment, but because there was no viable alternative. AI introduces that alternative, and it does so unevenly across these different layers of decision-making. As a result, what is changing is not simply the execution of work, but the distribution of decision authority within it.</p><h3>It is useful to think about this shift across three layers of decision.</h3><p><strong>At the first layer are task-level decisions,</strong> which are embedded within discrete activities and are characterized by clearly defined boundaries. These decisions operate within a constrained space of inputs, rules, and expected outputs, even when the underlying logic is highly sophisticated. Examples include classifying data, routing requests, generating initial drafts, reconciling records, or flagging anomalies.</p><p>It is important to be precise here. The defining feature of this layer is not simplicity, nor is it the absence of expertise. Many task-level decisions can be complex and may historically have required significant domain knowledge to perform well. However, they remain bounded in the sense that the decision space is constrained and the criteria for success can be specified with reasonable clarity. This distinction matters because boundedness, not difficulty, is what makes a decision amenable to automation. As a result, even highly skilled professionals operating primarily at this level are exposed. Expertise does not provide protection if it is consistently applied within a constrained decision space that can be modeled, learned, and replicated by AI.</p><p>At this layer, AI acts as an executor, and it is here that the most rapid and visible progress has already occurred.</p><p><strong>At the second layer are process-level decisions,</strong> which determine how work moves across activities, systems, and actors. These decisions involve sequencing, escalation, coordination, and adaptation in response to changing conditions. Rather than focusing on a single task, they shape the pathways through which work is completed. At this level, AI begins to function not just as an executor, but as an orchestrator, influencing the structure and flow of work itself.</p><p><strong>At the third layer are value-level decisions, </strong>which define what matters within a given context. These include judgments about priorities, acceptable trade-offs, risk tolerance, customer impact, and long-term consequences. Such decisions are inherently contextual and carry accountability, as they determine not just how work is done, but why it is done in a particular way. While AI can support these decisions through analysis, simulation, and pattern recognition, the responsibility for interpreting context and making consequential choices remains fundamentally human.</p><p>When viewed through this lens, the implications for individual careers become more precise. As AI absorbs decisions from the bottom up, work that is primarily composed of bounded, task-level decisions becomes increasingly automatable. Work that involves shaping and coordinating processes becomes increasingly augmented, as humans and AI interact to optimize flow and performance. Work that depends on contextual judgment and value-based decision-making becomes more, not less, important, because it sits at the point where accountability and consequence converge.</p><p>What is often underestimated in this transition is that exposure is not limited to routine or low-skill work. Any form of contribution that is expressed primarily through bounded decision-making, regardless of the level of expertise involved, is susceptible to automation. This reframes the risk entirely, shifting the focus from how difficult a task appears to how constrained the decision space actually is.</p><h3>This shift requires a different kind of clarity from professionals who want to remain relevant.</h3><p><strong>The first requirement is clarity about value.</strong> Rather than understanding your role as a collection of tasks or responsibilities, you need to understand it in terms of the value it ultimately produces. This involves looking beyond your immediate job description and situating your work within a broader structure that includes your level of contribution, your job family, your function, and the industry context in which you operate. From this perspective, you can identify what may be described as the last-mile deliverable, which is the outcome whose value persists regardless of how upstream activities are automated or reconfigured. Equally important is the ability to identify the variables that influence the quality and relevance of that deliverable, because those variables are where the most consequential decisions are made.</p><p><strong>The second requirement is clarity about your own mode of contribution.</strong> This goes beyond identifying skills or competencies and instead focuses on understanding the type of challenges that you are naturally inclined to engage with. It requires asking which kinds of problems energize you, which forms of work you find intrinsically meaningful, and which activities you would continue to pursue even if they could be partially or fully automated. As AI absorbs more routine and bounded work, the differentiating factor for individuals is not generic capability, but the ability to operate effectively in areas that require judgment, synthesis, and interpretation. Your core strength should align with these areas.</p><p><strong>The third requirement is clarity about fit within the value chain. </strong>It is not sufficient to understand value in the abstract or to have a well-developed sense of personal strengths in isolation. The critical step is to map your mode of contribution onto the specific points in the value chain where it has the greatest impact on the last-mile deliverable. This positioning determines not only your relevance, but also the degree to which your role is defensible as the structure of work evolves.</p><h3>Once these three forms of clarity are established, the role of AI can be approached more strategically.</h3><p><strong>At the level of task decisions,</strong> the objective should be to offload as much as possible. Allowing AI to handle bounded, repetitive decision-making reduces friction, increases speed, and compensates for areas where your own capabilities may be less developed or less aligned with your interests.</p><p><strong>At the level of process decisions,</strong> the relationship becomes more collaborative. AI can be used to design, test, and optimize workflows, to surface patterns that are not immediately visible, and to simulate alternative approaches. However, human oversight remains important in managing coordination complexity and interpreting contextual nuances that may not be fully captured in the data.</p><p><strong>At the level of value decisions, </strong>the role of the individual becomes more pronounced. While AI can provide analytical support, the responsibility for defining priorities, interpreting trade-offs, and making decisions with significant consequences cannot be fully delegated without introducing risk. This is where professional judgment, shaped by experience and context, becomes the primary source of value.</p><p>This distribution of responsibility across the three decision layers leads to a different approach to capability development. Rather than attempting to become broadly more skilled across all areas, individuals need to be more deliberate in how they invest in their development. Capabilities that are primarily associated with task-level decisions, and which are therefore more susceptible to automation, should receive less emphasis over time. Greater attention should be given to developing fluency in working with AI at the process level, particularly in understanding how to orchestrate and redesign work. Most importantly, individuals should deepen their capacity for value-level judgment, where the ability to interpret context, manage trade-offs, and take responsibility for outcomes becomes central.</p><p>In this context, the idea of “<em>keeping your job</em>” is somewhat misleading. What is actually required is a continuous repositioning within a changing system of work. This involves anchoring yourself to value, aligning your natural mode of contribution with that value, and deliberately delegating lower-order decisions to AI so that you can focus on higher-order judgment.</p><p>When approached in this way, AI does not simply represent a threat to employment. It becomes a mechanism for removing constraints, enabling individuals to operate more consistently in areas where their strengths have the greatest impact. This not only enhances effectiveness, but also creates the conditions for a form of professional experience that is both more meaningful and more sustainable over time.</p><h3>References and Intellectual Foundations</h3><p>This perspective builds on established work across decision-making, organizational theory, and learning:</p><ul><li>Herbert A. Simon (1960, 1977), <em>Administrative Behavior</em>; <em>The Sciences of the Artificial</em></li><li>Gary Klein (1998), <em>Sources of Power: How People Make Decisions</em></li><li>Daniel Kahneman (2011), <em>Thinking, Fast and Slow</em></li><li>Henry Mintzberg (1973), <em>The Nature of Managerial Work</em></li><li>James G. March (1994), <em>A Primer on Decision Making</em></li><li>Michael Porter (1985), <em>Competitive Advantage</em></li><li>Richard Rumelt (2011), <em>Good Strategy/Bad Strategy</em></li><li>Etienne Wenger (1998), <em>Communities of Practice</em></li><li>Erik Brynjolfsson and Andrew McAfee (2014, 2017), <em>The Second Machine Age</em>; <em>Machine, Platform, Crowd</em></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a20c921b602" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/keeping-your-job-in-the-age-of-ai-5a20c921b602">Keeping Your Job in the Age of AI</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Hierarchy Beyond Status: A Humanized View of Professional Progress]]></title>
            <link>https://medium.com/human-at-work/hierarchy-beyond-status-a-humanized-view-of-professional-progress-5f4e784001bb?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/5f4e784001bb</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[leadership-development]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[career-development]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Fri, 24 Apr 2026 07:40:09 GMT</pubDate>
            <atom:updated>2026-04-24T08:43:56.640Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QXwYy6e2rGCXcNZJozt4zA.png" /></figure><p>When most people think about career progress, they think about promotion. They think about moving up, earning more, gaining status, and sitting higher in a visible hierarchy. That is understandable. In most professional environments, hierarchy is the most obvious language through which progress is recognized. Titles rise, scope broadens, compensation increases, and the symbols of seniority become more visible.</p><p>But I have long felt that this way of thinking is too shallow to explain what professional progress really is.</p><p>It describes the surface of progression without explaining its substance. It tells us that something changes, but not what actually changes in the nature of the contribution being expected. If people at different levels are compensated differently, trusted differently, and held differently, then the real question is not simply who is above whom. The real question is what kind of value a person is now expected to carry, protect, enable, or deliver.</p><p>Without that question, hierarchy becomes easy to misunderstand. Some people chase it as though elevation itself were proof of growth. Others resist it as though it were merely a system for ranking human beings. Both responses are understandable, and both miss something important. Hierarchy, when it is legitimate, is not fundamentally about superiority. It is about differentiated responsibility. It is about what changes when a role requires broader perspective, more consequential judgment, greater decision rights, a different relationship to execution, and a more developed human capacity to carry all of that well.</p><p>My own view is that professional hierarchy makes the most sense when it is understood not as rank, but as progression in value responsibility. What changes from one level to another is not merely one’s position relative to other people. What changes is the breadth of perspective one must hold, the results one is accountable for, the authority one requires in order to act meaningfully on those results, the mode through which one contributes to value creation, and the capability needed to do so reliably. Seen this way, hierarchy becomes less about standing above others and more about carrying a different order of burden.</p><p>That distinction matters because it restores a human meaning to progress. It shifts the conversation away from corporate ladders and toward the actual structure of contribution. It helps explain why higher levels are not simply “more important” versions of lower levels, but different forms of work with different demands, different risks, and different consequences. It also helps explain why career progress is often so uneven. Many people do not struggle at a new level because they are lazy or unintelligent. They struggle because they interpret the new level through the logic of the old one.</p><h3>The first thing that changes across levels is perspective.</h3><p>At more immediate levels of work, success often depends on closeness. The person is near the task, near the customer, near the process, near the materials, near the real-time conditions under which value is actually produced. They can see breakdowns directly. They can sense nuance through contact. They often learn through doing, correcting, adjusting, and observing the direct consequences of action. This kind of nearness matters because the ground is not a secondary place in professional life. It is where much of the real work happens.</p><p>As one progresses, however, broader patterns begin to matter more. The person can no longer focus only on what is immediately in front of them. They must think across interdependencies, across functions, across time horizons, across trade-offs, across competing priorities, and across the larger conditions under which work succeeds or fails. The work increasingly requires abstraction. Not because abstraction is inherently superior, but because wider responsibility demands a wider field of view.</p><p>Yet there is a danger built into that widening. The broader the vantage point, the greater the risk of distance from lived operational reality. Reports begin to stand in for observation. Metrics begin to stand in for meaning. Dashboards begin to stand in for the texture of the work itself. The person may become more informed in one sense and less grounded in another. This is one of the quiet tensions of hierarchy. Higher-level work requires abstraction, but good higher-level work also requires disciplined return to the ground. It requires intellectual altitude without epistemic detachment. It requires the ability to think broadly without becoming seduced by representations of reality that no longer correspond to what is actually happening where value is delivered.</p><h3>The second thing that changes is accountability.</h3><p>In everyday organizational language, responsibility is often treated as if it were one thing. It is not. There is a profound difference between being responsible for doing work and being accountable for whether a broader result is achieved. At more immediate levels, success may be judged substantially through completion, quality, timeliness, and local effectiveness. Those are serious demands. But as professional scope expands, the person’s relationship to outcomes changes. They are no longer only responsible for performing activity well. They are increasingly answerable for whether the intended consequence of that activity is realized.</p><p>That shift changes the nature of the work. The work becomes more decisional, not in the sense that execution disappears, but in the sense that results are increasingly shaped by judgment. The person must interpret context, define priorities, decide what matters most, determine what trade-offs are acceptable, and act under conditions where certainty is incomplete and consequences are real. This is why progression is not simply a matter of having more to do. It is a matter of becoming answerable for a broader order of result.</p><p>This distinction is important because <em>many professionals continue to judge themselves by effort long after their role has begun to judge them by consequence. </em>They work hard, stay busy, move quickly, and remain deeply engaged, but the level now requires something more than visible effort. It requires the capacity to hold a result together. It requires the ability to make sense of competing demands and decide which consequences are worth absorbing in order to protect the outcome that matters most. At that point, busyness may even become a liability. A person can remain very active while failing to carry the actual accountability of the level.</p><h3>The third thing that changes is authority.</h3><p>This is one of the most practical and one of the most neglected aspects of professional hierarchy. If a person is genuinely accountable for a result, they must possess the authority required to shape that result in some meaningful way. Otherwise, what exists is not accountability in the proper sense, but a theatrical imitation of it. The person is given the burden of consequence without the power of action.</p><p>Organizations do this all the time. They tell people to “own” outcomes while retaining critical decisions elsewhere. They hold someone answerable for delivery while depriving them of the ability to allocate resources, change priorities, challenge constraints, or intervene across boundaries. They assign strategic responsibility without granting strategic discretion. Then, when results suffer, they describe the problem as one of leadership, ownership, or mindset when the deeper issue is structural incoherence.</p><p>A humanized view of hierarchy does not allow that confusion to go unchallenged. The broader and more consequential the result a person is expected to carry, the more real autonomy, discretion, and legitimized influence that person must have. Authority does not have to mean unilateral power, and it certainly does not mean ego. But it must mean enough decision latitude to make accountability genuine. Without that alignment, hierarchy becomes less a structure of differentiated contribution and more a system for distributing blame downward while centralizing control upward.</p><h3>The fourth thing that changes is the mode of contribution.</h3><p>At more immediate levels, contribution is often direct. The person does the work. They produce the output, solve the problem, deliver the service, handle the case, close the loop, fix the issue, build the thing, or complete the analysis. Their relationship to value creation is visible and immediate. They are close to the point where effort becomes outcome in concrete form.</p><p>As scope broadens, however, contribution increasingly becomes indirect. The person contributes by shaping the conditions under which value can be produced more reliably across a wider field. They may coordinate effort rather than execute every part of it. They may remove constraints rather than solve each operational problem personally. They may design standards, allocate attention, set direction, reconcile tensions, build capability in others, protect capacity, or intervene only where judgment can unlock disproportionate value. In other words, the contribution becomes less about doing all the important work oneself and more about ensuring that important work can be done well by the wider system.</p><p>This is often where people become confused about hierarchy because indirect contribution can look less “real” than direct contribution. It may seem as though the person further from the ground is contributing less because they are touching fewer immediate tasks. Sometimes that is true. Some people do hide behind hierarchy. But when hierarchy is functioning properly, indirect contribution is not the absence of work. It is a different form of work. It is work aimed at coherence, reliability, prioritization, integration, protection, and direction. It is work that helps many other pieces of work succeed together.</p><p><em>This is also why hierarchy should never be interpreted as a moral ranking of dignity. Ground-level work is not lesser work. </em>In many cases it is closer to the actual moment where value is delivered. The point is not that higher-level work matters more in some human sense. The point is that it is differently related to value creation. One role may deliver value directly. Another may enable, coordinate, and protect the conditions under which that value can be delivered consistently and at greater scale. Both matter. They are simply not the same.</p><h3>The fifth and most human shift is capability.</h3><p>A new level of work does not merely ask for more effort from the same person. It often asks for a different configuration of perception, judgment, self-management, knowledge, behavior, and developmental maturity. This is why formal promotion so often fails to produce real progression. The title changes. The salary changes. The expectations change. But the person may still be operating with the success logic of the previous level.</p><p>Someone who succeeded by being the most technically reliable person in the room may now need to succeed by growing the capability of others instead of proving their own. Someone who succeeded through personal responsiveness and visible activity may now need to succeed through selectivity, prioritization, and disciplined non-interference. Someone who built credibility through having answers may now need to function well in ambiguity, where holding the right questions matters more than immediately supplying solutions. These are not small adjustments. They are shifts in the human being the role requires.</p><p>That is why capability cannot be reduced to skill alone. It includes the willingness to reinterpret success. It includes the maturity to accept that past strengths may become future traps when the context changes. It includes the honesty to recognize when the demands of the level exceed one’s current operating assumptions. And it includes a developmental posture toward that gap. Professional progress is therefore never purely external. It is not just about carrying more. It is about becoming capable of carrying differently.</p><p>This, to me, is one of the most important reasons to humanize our understanding of hierarchy. Too often, people are promoted as though scope were simply added to them. But progression is not additive in that simple sense. It is often transformative. It requires a partial reorganization of one’s own professional identity. The role begins to ask different things of attention, temperament, timing, restraint, confidence, and judgment. If we do not understand that, we misdiagnose performance struggles and design weak development responses. We assume people need more effort when they may need a different posture. We assume they need more knowledge when they may need better judgment. We assume they need confidence when they may actually need humility, or vice versa.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/405/1*ob75nhIuPxPOwGqicivwgQ.png" /></figure><p>Once hierarchy is understood in these terms, it also becomes clear that it extends beyond the formal organization chart.</p><p>The same logic appears whether someone is an employee, contractor, entrepreneur, founder, consultant, freelancer, or small business owner. The issue is not title. The issue is the level of value responsibility actually being carried. A founder can still operate from a narrow, overly immediate horizon. A senior executive can still lack meaningful authority in a highly politicized structure. A specialist with no managerial title can carry immense consequence within a complex domain. A consultant may influence strategic decisions without ever appearing inside the client’s hierarchy. The underlying question is the same in each case: what order of contribution, consequence, and judgment does this role or position actually require?</p><p>That broader applicability matters because it frees the idea of hierarchy from a narrow corporate frame. It allows us to see that progress is not simply about moving upward inside an institution. It is about moving into broader, deeper, or more consequential forms of responsibility, regardless of the contractual or structural arrangement within which that responsibility sits. It is about the increasing scale of what one must see, hold, decide, enable, and become.</p><p>That, for me, is the more human meaning of hierarchy.</p><p>It is not a justification for vanity. It is not a defense of status inflation. It is not an attempt to romanticize being further from the ground. It is a way of making professional progress intelligible in terms of value, consequence, and human development. It is a way of asking what really changes when a person moves from one level of work to another.</p><p>Seen this way, hierarchy is not simply a ladder of rewards. It is a structure of differentiated demands. It asks different things of perception, different things of judgment, different things of authority, different things of contribution, and different things of the self. It is therefore best understood not as a ranking of persons, but as a progression in the scale and nature of responsibility one is expected to carry well.</p><p>And perhaps that is the question that matters most. Not whether one has moved up, but whether one is becoming equal to the level of value responsibility one now claims or has been asked to hold.</p><ol><li>What must I now be able to see?</li><li>What results am I now answerable for?</li><li>Do I have the authority required to carry them honestly?</li><li>How has my mode of contribution changed?</li><li>And am I becoming the person this level of work now requires?</li></ol><p>Those questions, more than title alone, are what make hierarchy meaningful.</p><h3>How this perspective builds on existing work</h3><p><em>This essay does not present differentiated levels of work as a new idea. It builds on established thinking about levels of responsibility and time-span in organizational roles, the actual nature of managerial work, bounded rationality and decision-making under constraint, organizational information processing, and reflective professional practice. Its contribution is to synthesize those strands into a more human-centered interpretation of professional progress, framed around changing value responsibility rather than status alone.</em></p><h3>Selected foundations</h3><ul><li>Elliott Jaques on levels of work and time-span of discretion.</li><li>Henry Mintzberg on the real character of managerial work.</li><li>Herbert Simon on bounded rationality and decision-making limits.</li><li>Jay Galbraith on organizations as information-processing systems under uncertainty</li><li>Donald Schön on reflective professional practice.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5f4e784001bb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/hierarchy-beyond-status-a-humanized-view-of-professional-progress-5f4e784001bb">Hierarchy Beyond Status: A Humanized View of Professional Progress</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Technologist Mind: What EUCs Tell Us About Who Should Be Shaping AI Transformation]]></title>
            <link>https://medium.com/human-at-work/the-technologist-mind-what-eucs-tell-us-about-who-should-be-shaping-ai-transformation-31b3f93e4110?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/31b3f93e4110</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Fri, 10 Apr 2026 16:43:45 GMT</pubDate>
            <atom:updated>2026-04-10T16:43:45.895Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vIiRn_VqV5fRuSCPY9rtHg.png" /></figure><p>A few years ago, actually, quite a while ago, I was working on a project that sought to reduce, ideally eliminate , End-User Computing applications. EUCs. If you’ve worked in any large organization, you know what these are: the Excel workbooks with macros that run entire reporting cycles, the Access databases that somehow became mission-critical, the clever combinations of Office tools held together by someone’s ingenuity and a few hundred lines of VBA.</p><p>The brief was clear. These things posed risks — to security, to process control, to data integrity. They sat outside IT governance. They couldn’t be audited properly. They made process improvement work nearly impossible because half the actual process lived inside someone’s personal file on a shared drive.</p><p>All of that was true. But there was something else in the room that nobody was paying enough attention to.</p><h3>The behavior behind the tool</h3><p>Organizations did ask why EUCs existed. The answers were usually known ; the main system didn’t cover a particular need, a process had a gap, and a workaround had become necessary. But the goal of the exercise typically outweighed those answers. The assumption was that the main system was the right system, and that the EUC represented the team’s unwillingness or inability to fully adopt the improved process the technology was designed to support. The focus stayed on the tool , how to bring people back to the sanctioned platform, how to close the gap in the system, and how to eliminate the workaround.</p><p>What got less attention was the behavior itself.</p><p>Someone in that team , or in a neighboring team , had looked at a process that wasn’t working, understood the friction, knew enough about the tools available to them to build something, and then built it. They didn’t submit a business case. They didn’t wait for IT. They saw a real need, had the means to address it, and acted. The need, the means, and the capability all existed in the same place. The EUC was born not from a purchase decision that required change management and adoption after the fact, but from a lived, immediate experience of the work.</p><p>That behavior , the agency, the technical fluency, the instinct to solve , is what I want to talk about. Because it’s the same behavior that should now be driving AI adoption from the inside, and it’s still not getting the attention it deserves.</p><h3>How we responded and what the response missed</h3><p>When organizations responded to EUCs, they generally did the right things. They sought to make EUCs visible, classify them by risk, and manage that risk; ideally, by upgrading the process so the main system could absorb what the EUC was doing, and the workaround would no longer be needed. That’s sound governance. I’m not criticizing the approach.</p><p>But the framing told a particular story. EUCs were positioned primarily as a risk and compliance issue, something to be managed, contained, and, where possible, retired. The people who built them were understood as users who had drifted from the sanctioned process, not as practitioners who had demonstrated something valuable.</p><p>And when the response focused on the tool and the process without equally attending to the behavior, something predictable happened. The EUCs that were eliminated without the underlying need being fully resolved didn’t always stay eliminated. People found other ways, sometimes less visible, sometimes with higher risk, because the friction that produced the original EUC was still there. The behavior didn’t stop. It just moved.</p><p>This isn’t a failure of governance. It’s a signal that the behavior itself was carrying information the organization needed , and that information was about more than just a process gap.</p><h3>The behavior is the asset</h3><p>When I look back at the EUC phenomenon, the thing that stands out most isn’t the tools people used or the risks those tools created. It’s what the behavior tells us about the people who exhibited it.</p><p>These were practitioners — not IT professionals by title, but people whose domain knowledge and problem-solving instinct converged with enough technical fluency to produce something useful. They understood the work. They felt the friction. They knew what the output needed to look like. And they had the agency to act on all of that without waiting for someone else to solve it for them.</p><p>That’s a particular kind of mind. I think of it as a technologist mind — not because these people were writing enterprise software, but because they instinctively reached for the tools available to them and shaped those tools to fit the work. They were the ones who knew Excel well enough to write a macro, who understood Access well enough to build a small database, who could see a process failing and had the initiative to bridge the gap.</p><p>And here’s what’s easy to overlook: these people typically maintained what they built. They debugged it. They cared about its accuracy because it directly affected their own work and the work of the people around them. That’s a form of accountability that many enterprise-deployed systems never achieve. The builder knew the logic, understood the edge cases, and felt the consequences of failure.</p><p>The risks of EUCs were real and needed to be governed. But the behavior that produced them, the agency, the fluency, the ownership, that was never the problem. That was an asset hiding in plain sight.</p><h3>The same behavior, new tools</h3><p>I’ve written before about the three levels at which AI is transforming work: task, process, and role. What I want to name here is that the EUC behavior could , and should , repeat at all three levels. But right now, it’s only really supported at one.</p><p>At the task level, it’s already possible because most organizations have some form of AI Chat implemented — Microsoft Copilot for most, Anthropic Claude or Google Gemini for an increasing number, with Anthropic Cowork slowly seeping in. Most people are also using one of these at home, just like they do with MS Office applications. So now, at the desk, someone uses ChatGPT or Claude to draft a report, clean data, or generate a formula. This is the equivalent of the quick Excel macro , a point solution for a specific friction. It happens in a chat window, it’s invisible to the organization, and it works. The person didn’t ask permission. They had a need, the tool was available, and they solved the problem. This is where most people’s experience of AI sits today, and it’s happening for the same reasons EUCs happened . The behavior is natural when the tools are accessible, and the need is real.</p><p>At the process level, the potential is enormous but the tools aren’t as accessible yet. This is where someone would build something more substantial , an automated workflow, a small application, a tool that reshapes how an entire sequence of work gets done. The EUC equivalent is the Access database that became load-bearing. But unlike Excel macros, this level of AI work requires more than a chat interface ; it requires what looks more like application development, but using natural language programming or similar no-code or low-code interfaces. The person isn’t just prompting anymore , they’re designing. These capabilities are emerging, but they’re not yet as universally available as a chat window. The people with the instinct to build at this level are there. The availability of the tools at work is catching up.</p><p>At the role level, we’re talking about something that’s still largely aspirational for most organizations. This is where someone would reimagine what their job actually is , building AI agents that handle recurring decisions, monitor conditions, or execute judgment-laden tasks that used to consume their day. The EUC equivalent is the person who didn’t just build a macro but fundamentally restructured how their function operated. This requires agentic platforms that are still maturing, and most organizations haven’t yet created the conditions for this kind of exploration.</p><p>But here’s the thing: the behavior that would drive all three levels is the same behavior that produced EUCs. The same ingredients , a real need, technical fluency, and the agency to act. What’s different is that the tools at the process and role levels aren’t yet as available as Excel once was. If the people who exhibit this behavior are recognized and empowered, if organizations actively put these emerging tools in their hands , the same pattern that produced EUCs could produce something far more valuable.</p><h3>Empowering the behavior</h3><p>This is the part I care about most, and where I think the real opportunity sits.</p><p>EUCs happened because Excel and Access were already on everyone’s desktop. The tools were available. The needs were real. Capable people connected the two. Nobody planned it. Nobody resourced it. It just happened because the conditions were right.</p><p>At the task level, those conditions are forming again, AI chat tools are landing on desktops and people are already using them the way they once used Excel. But at the process and role levels, the conditions aren’t there yet. The tools are emerging, not universal. And unlike with EUCs, where the behavior simply happened because the means were at hand, the next two levels won’t happen on their own. They need to be enabled.</p><p>That’s the opportunity. We know what this behavior looks like. We’ve seen it before. We know that inside every organization, there are people with the technologist mind — the ones who built the EUCs, who bridged the gaps, who solved problems with whatever was available to them. They’re still there. And the instinct hasn’t changed.</p><p>So the question isn’t whether to allow it. The question is whether to recognize it, resource it, and direct it.</p><p>These people have a specific and valuable combination: they understand the work deeply, they have enough technical fluency to bridge the gap between need and solution, and they have the initiative to act without waiting for permission. That profile doesn’t just happen to be useful for AI adoption. It’s the <em>most</em> useful profile for AI adoption. Because AI adoption that matters doesn’t start with a platform purchase or an enterprise license. It starts with someone who knows where the friction is and has the agency to do something about it.</p><p>Organizations should be identifying these people across roles, teams, and departments. Not to put them in an AI Center of Excellence where their instincts get bureaucratized, but to give them better tools, clearer guardrails, and the organizational permission to do what they were already doing, except now with AI-grade leverage. Give them the platforms. Give them the sandbox environments. Give them the frameworks that say “yes, and here’s how” instead of “no, because risk.” Let them surface the quick, high-value opportunities for AI to absorb friction in tasks, processes, and roles, not from a strategy deck, but from the lived reality of doing the work.</p><p>The governance will still matter. Making AI usage visible, classifying it by risk, and managing that risk proportionally are all necessary, just as they were with EUCs. But governance is the how. The behavior is the what. And if we start with the behavior, if we recognize it as the asset it is and build around it, the governance becomes something that enables rather than constrains.</p><h3>What this is really about</h3><p>I keep coming back to the same idea across everything I write about work and AI: the people closest to the work are the ones who understand it best. Not the people designing the systems. Not the people writing the policies. The people doing the work.</p><p>EUCs were proof of that. They were imperfect, ungoverned, sometimes risky, and they were born from a deeper understanding of the work than most formal systems ever achieved. The people who built them weren’t drifting from the process. They were compensating for its limits. They were doing what humans do when the tools don’t fit : they improvise, they build, they find a way.</p><p>In the age of AI, that same instinct, that same agency, that same technologist mind that lives inside practitioners across every role, team, and department , that’s not a risk to be managed. It’s the most valuable thing an organization has.</p><p>The question is whether we’ve learned enough from the EUC story to treat it that way.</p><p><em>This builds on my ongoing exploration of AI, work, and human agency , particularly the three levels of AI transformation (task, process, role) and the idea that the future of work depends not on what AI can do, but on who gets to shape how it’s used.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=31b3f93e4110" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/the-technologist-mind-what-eucs-tell-us-about-who-should-be-shaping-ai-transformation-31b3f93e4110">The Technologist Mind: What EUCs Tell Us About Who Should Be Shaping AI Transformation</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Epistemic Contact: Meaning, Judgment, and the Structural Consequences of Organisational Scale]]></title>
            <link>https://medium.com/human-at-work/epistemic-contact-meaning-judgment-and-the-structural-consequences-of-organisational-scale-a95bc8bcde70?source=rss-7b135f973c88------2</link>
            <guid isPermaLink="false">https://medium.com/p/a95bc8bcde70</guid>
            <category><![CDATA[practitioner-intelligence]]></category>
            <category><![CDATA[epistemic-contact]]></category>
            <category><![CDATA[leadership-development]]></category>
            <category><![CDATA[management-and-leadership]]></category>
            <dc:creator><![CDATA[Dehumo Bickersteth]]></dc:creator>
            <pubDate>Mon, 06 Apr 2026 06:29:24 GMT</pubDate>
            <atom:updated>2026-04-06T06:29:24.900Z</atom:updated>
            <content:encoded><![CDATA[<p>Organisational scaling requires the distribution of work across specialised roles and hierarchical layers. While this distribution is well understood as a coordination problem, its epistemic consequences remain undertheorised. This paper introduces <strong>epistemic contact</strong> as a structural condition for organisational legitimacy and effectiveness, defined as the degree to which any organisational role maintains a live, consequential, reality-disciplined relationship with the substance of the work it mediates. Drawing on scholarship in alienation theory, organisational sensemaking, expertise research, and the philosophy of work, the paper develops a linked analysis of how organisational scaling produces four distinct epistemic separations — epistemic, executive, material, and consequential — that progressively disconnect decision-makers from the work their decisions affect. The paper identifies the <strong>aspiration gap</strong> as the central mechanism through which this disconnection is culturally reproduced: career architectures that define advancement as progressive distance from practice, thereby teaching organisational members to aspire to the very conditions that degrade their judgment and erode their meaning. Building on the Knowing–Working–Doing (KWD) framework and the concept of reflexive competence (Bickersteth, 2025), the paper extends these constructs from the individual to the organisational level, proposing <strong>organisational reflexive competence</strong> as the capacity of an organisation’s layers to maintain awareness of the work they mediate without collapsing into it or drifting away from it. Implications are drawn for career architecture, role design, learning systems, organisational evaluation, and AI strategy.</p><p><strong>Keywords:</strong> Epistemic contact, aspiration gap, organisational scale, meaningful work, Knowing–Working–Doing framework, organisational reflexive competence, expertise, career architecture, AI and organisational design</p><p><strong>Dehumo Bickersteth</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K7Y3YtvN9nvnSjcrveYkpQ.png" /></figure><h3>1. Introduction: The Distance Nobody Designed</h3><p>Most organisations are structured so that the people most responsible for outcomes are the least directly connected to the work that produces them.</p><p>This observation is not novel. Marx (1844) identified the alienation of the worker from the product, the process, and their own productive activity. Arendt (1958) distinguished labour, work, and action as fundamentally different modes of human engagement with the world. Mintzberg (1973, 2004) documented the disconnect between what managers actually do and the rationalist models that claim to describe their work. Weick (1995) showed that organisational sensemaking degrades as distance from the action increases. Each of these contributions illuminates a dimension of the problem. None, I will argue, provides a sufficient account of the structural mechanism through which organisational scaling produces epistemic disconnection, nor of the cultural process through which that disconnection is reproduced and even desired.</p><p>This paper addresses both gaps. It introduces <strong>epistemic contact</strong> as a formal construct — a structural condition that can be designed for or designed against, measured, and evaluated — and identifies the <strong>aspiration gap</strong> as the primary mechanism through which organisations culturally reproduce the erosion of epistemic contact across their hierarchical layers.</p><p>The paper proceeds as follows. Section 2 establishes the theoretical foundations, drawing on traditions in alienation theory, organisational learning, expertise research, and the philosophy of work. Section 3 introduces the four-dimensional circuit of meaningful work and traces how organisational scaling opens it along four distinct axes. Section 4 formally defines epistemic contact and distinguishes it from adjacent constructs. Section 5 develops the aspiration gap as the paper’s central diagnostic contribution. Section 6 extends the concepts of the Knowing–Working–Doing framework and reflexive competence (Bickersteth, 2025) from the individual to the organisational level. Section 7 draws implications for organisational design. Section 8 concludes with testable propositions.</p><h3>2. Theoretical Foundations</h3><p>The problem of how organisational structure mediates the relationship between people and work sits at the convergence of several established scholarly traditions.</p><h3>2.1 Alienation and the Separation of Worker from Work</h3><p>Marx’s (1844) early manuscripts identify four forms of alienation produced by industrial capitalism: alienation from the product of labour, from the act of production, from species-being (the human capacity for free, conscious, creative activity), and from other workers. While Marx’s analysis was situated within a specific political-economic framework, the structural insight — that the organisation of production can systematically sever the relationship between the worker and the conditions under which work generates meaning — transcends its original context.</p><p>Arendt (1958) drew a complementary but distinct set of distinctions. Her tripartite framework of labour (the metabolic cycle of biological necessity), work (the fabrication of durable objects that constitute a shared human world), and action (the initiation of something new in the presence of others) identifies different modes of human engagement, each with its own relationship to meaning. Arendt’s concern was that modernity had collapsed these distinctions, reducing work to labour and marginalising action. The present paper shares Arendt’s structural orientation but focuses on a different collapse: the reduction of organisational roles to information processing, which marginalises the direct engagement with the substance of work that both meaning and judgment require.</p><h3>2.2 Organisational Learning and Sensemaking</h3><p>Argyris and Schön (1978) established that organisations learn through the detection and correction of error, and that this learning can operate at two levels: single-loop learning (correcting actions within existing frameworks) and double-loop learning (questioning the frameworks themselves). Their insight that organisations frequently exhibit a gap between espoused theory and theory-in-use — between what members say they do and what they actually do — is directly relevant to the epistemic contact problem: the gap between what upper-layer decision-makers believe about the work and what the work actually involves is a specific instance of the espoused-theory/theory-in-use divergence.</p><p>Weick (1995) demonstrated that organisational sensemaking — the process through which members construct shared understandings of ambiguous events — is shaped by proximity to the action. Members closer to the work construct meanings grounded in direct experience; members further from the work construct meanings from representations that may diverge significantly from the experience they purport to describe. Weick’s analysis of the Mann Gulch disaster (1993) showed how the collapse of sensemaking under crisis conditions was related to the breakdown of the connection between organisational roles and the reality they were designed to manage.</p><p>Senge (1990) argued that learning organisations require what he called “systems thinking” — the capacity to perceive the whole system rather than its parts in isolation. While Senge’s contribution focused primarily on cognitive frameworks, the present paper argues that systems thinking without epistemic contact produces sophisticated but ungrounded models — frameworks that are internally coherent but disconnected from the operational reality they claim to represent.</p><h3>2.3 Expertise and Situated Judgment</h3><p>The expertise literature provides the epistemic foundation for the argument. Polanyi (1966) demonstrated that expert knowledge is substantially tacit — irreducible to explicit rules or verbal formulations. Schön (1983) showed that professional competence operates through knowing-in-action and reflection-in-action, both of which depend on direct engagement with the situations in which judgment is exercised. Dreyfus and Dreyfus (1986) traced the developmental trajectory from rule-following novice to intuitive expert, demonstrating that the highest levels of performance are precisely those least amenable to explicit articulation.</p><p>Klein (1998) provided empirical grounding through his research on naturalistic decision-making, showing that experienced practitioners in high-stakes environments employ recognition-primed decisions — pattern-based judgments that draw on extensive experience rather than on deliberate analysis. Eraut (2000) demonstrated that professional learning is substantially non-formal, implicit, and embedded in practice, and therefore invisible to organisational systems that require articulation as evidence of competence.</p><p>The critical implication of this literature for the present paper is that the kind of understanding on which sound organisational judgment depends — situated, tacit, consequence-tested, perceptually grounded — is developed and maintained through direct engagement with the work, and degrades when that engagement is severed. This is the epistemic foundation of the contact requirement.</p><h3>2.4 Meaningful Work</h3><p>Research on meaningful work provides the motivational and existential dimension. Rosso, Dekas, and Wrzesniewski (2010) identified self-efficacy, purpose, authenticity, and belongingness as primary mechanisms through which work becomes meaningful — mechanisms that depend substantially on the conditions the closed circuit provides: visible outcomes, causal connection between effort and result, and alignment between the person’s capacities and the work’s demands. Lips-Wiersma and Morris (2009) argued that meaningful work requires a dynamic balance among developing the self, serving others, expressing full potential, and unity with others — a balance that organisational structures can enable or undermine.</p><p>Csikszentmihalyi (1990) demonstrated that the psychological state of flow — characterised by deep absorption, clear goals, immediate feedback, and a match between skill and challenge — is the condition under which work is most intrinsically rewarding. The structural parallels between flow and the closed circuit described in this paper are not accidental: both depend on the integration of understanding, action, and consequence in a single agent.</p><p>Sennett (1998, 2008) and Crawford (2009) argued, from sociological and philosophical perspectives respectively, that modern economic arrangements have degraded the conditions under which work can serve as a source of identity, skill development, and meaning. Sennett focused on the corrosion of character produced by flexible capitalism; Crawford focused on the cognitive richness of manual work relative to the impoverishment of much knowledge work. The present paper builds on both but shifts the analytical focus from occupational categories to organisational structure — arguing that it is the structure of the relationship between person and work, not the nature of the work itself, that determines whether meaning is generated.</p><h3>2.5 Organisational Structure and Managerial Work</h3><p>Mintzberg (1973) demonstrated through empirical observation that managerial work bears little resemblance to the rational planning, organising, and controlling functions described by classical management theory. Managers work in short bursts, respond to interruptions, communicate informally, and maintain understanding through direct contact rather than through formal information systems. Mintzberg’s later work (2004) argued that management education had become disconnected from management practice — producing graduates fluent in analytical frameworks but lacking the situated judgment that effective management requires.</p><p>Chandler (1962) documented the co-evolution of organisational strategy and structure, showing how the demands of scale and diversification produced the multidivisional form that characterises large modern corporations. Weber’s (1922/1978) analysis of bureaucratic rationalisation explained how the pursuit of efficiency and predictability produces hierarchical structures that systematically replace individual judgment with procedural compliance. Both analyses describe the structural trajectory that this paper examines: the progressive layering of organisational roles, each solving a coordination problem while introducing an epistemic separation.</p><h3>3. The Four-Dimensional Circuit and Its Opening</h3><h3>3.1 The Circuit of Meaningful Work</h3><p>I propose that work generates meaning, develops judgment, and sustains engagement when four dimensions are integrated in the person performing it:</p><p>The <strong>epistemic dimension</strong>: the person understands the work — not necessarily in theoretical terms, but in terms sufficient to guide situated judgment. They can perceive the work’s patterns, constraints, and possibilities directly.</p><p>The <strong>executive dimension</strong>: the person performs the work directly, encountering its resistances and receiving immediate feedback on the quality of their decisions.</p><p>The <strong>material dimension</strong>: the person controls the means through which the work is done — the tools, systems, resources, and conditions of production.</p><p>The <strong>consequential dimension</strong>: the person experiences the outcomes of the work — not as data or reports, but as lived realities that affect their own condition.</p><p>When these four dimensions are held together — what I call the <em>closed circuit</em> — the person’s relationship to the work produces meaning (through causal connection between effort and outcome), judgment (through the progressive refinement of perception under consequence), accountability (through direct exposure to the effects of one’s decisions), and identity (through the anchoring of self-understanding in the activity of the work).</p><p>The integration of these dimensions is not incidental. It is maintained by what I identify as the <em>binding agent</em> of the circuit: ownership — understood not in the narrow legal sense but as the structural condition of having a stake that integrates understanding, action, and consequence. When the person who does the work also controls the means and bears the consequences, the four dimensions are held in alignment by the convergence of epistemic, practical, and material interests.</p><h3>3.2 The Four Separations</h3><p>Organisational scaling opens the circuit along each of the four dimensions:</p><p><strong>Epistemic separation</strong> occurs as work is divided and distributed, so that no single person comprehends the whole. The practitioner understands their domain; the manager understands the coordination structure; the leader understands the strategic context. Each holds a partial understanding. None holds the integrated understanding that the closed circuit produces.</p><p><strong>Executive separation</strong> occurs as coordination roles emerge. The person who decides what should be done and the person who does it become different people. The feedback that would discipline the decision-maker’s judgment — the direct encounter with the work’s resistances — is filtered through reports and representations.</p><p><strong>Material separation</strong> occurs as ownership of tools, systems, and resources migrates to those who do not use them directly. Practitioners work with means they did not choose, in environments they did not design, under constraints set by others whose optimisation logic may diverge from the logic of effective practice.</p><p><strong>Consequential separation</strong> occurs as insulating layers develop between decision-makers and the effects of their decisions. The people who make the most consequential decisions experience the least direct feedback from those decisions’ effects.</p><p>Each separation solves a legitimate coordination problem. Together, they produce an organisation in which no single layer holds the full circuit — and in which the meaning-generating, judgment-developing properties of the circuit are distributed across roles that may never directly interact.</p><h3>3.3 Epistemic Drift</h3><p>I define <strong>epistemic drift</strong> as the progressive divergence between the understanding of work held at upper organisational layers and the reality of work as experienced at the point of delivery.</p><p>Epistemic drift is incremental, largely invisible to those experiencing it, and structurally produced rather than individually caused. It results from the compounding of the four separations described above: as decision-makers lose direct contact with the work’s epistemic, executive, material, and consequential dimensions, their models of the work become increasingly constructed from representations rather than from direct engagement. These representations — reports, metrics, dashboards, narratives — are systematically impoverished relative to the reality they describe, because each layer of translation compresses texture, filters nuance, and suppresses the specific resistances that would calibrate the decision-maker’s judgment.</p><p>Epistemic drift produces identifiable patterns of organisational dysfunction: metric fixation (Muller, 2018), in which proxy measures are optimised at the expense of the outcomes they represent; normalised deviance (Vaughan, 1996), in which the boundaries of acceptable risk shift incrementally as decision-makers are insulated from the consequences of their risk assessments; and the confidence-competence gap, in which decision-makers’ subjective confidence in their understanding of the work remains high even as the objective accuracy of that understanding deteriorates.</p><h3>4. Epistemic Contact: Definition and Distinctions</h3><h3>4.1 Formal Definition</h3><p><strong>Epistemic contact</strong> is defined as the degree to which someone in any organisational role maintains a live, consequential, reality-disciplined relationship with the substance of the work their role exists to enable.</p><p>This definition synthesises key elements from the traditions reviewed above. “Live” reflects the expertise literature’s emphasis on the currency of situated knowledge — knowledge that is developed and maintained through ongoing engagement, not stored from past experience (Klein, 1998; Eraut, 2000). “Consequential” reflects the decision-making literature’s emphasis on accountability and skin in the game as conditions for sound judgment (Taleb, 2012; Simon, 1956). “Reality-disciplined” reflects the sensemaking literature’s distinction between contact with events and contact with representations of events (Weick, 1995). “The substance of the work” specifies that the contact must be with the organisation’s productive activity, not merely with its processes, politics, or culture.</p><p>Epistemic contact is not a binary condition. It is a continuous variable — a matter of degree that can be assessed, designed for, and measured. An organisational role can have more or less epistemic contact with the work it mediates, and the degree of contact will affect the quality of judgment exercised in that role.</p><h3>4.2 Distinguishing Epistemic Contact from Adjacent Constructs</h3><p>Epistemic contact must be distinguished from several concepts with which it shares surface resemblance.</p><p><strong>Epistemic contact is not information access.</strong> A decision-maker who receives comprehensive reports, reviews detailed dashboards, and reads every metric may be well-informed but does not thereby possess epistemic contact. Information is a representation of reality; epistemic contact is engagement with reality prior to its compression into information. The distinction parallels Polanyi’s (1966) observation that tacit knowledge — the kind developed through direct engagement — is fundamentally different from the explicit knowledge that information systems transmit.</p><p><strong>Epistemic contact is not micromanagement.</strong> Micromanagement represents the collapse of organisational differentiation — the manager who does the work rather than coordinating it. Epistemic contact does not require the decision-maker to perform the work. It requires them to maintain sufficient direct engagement with the work to calibrate their judgment about it. The distinction is between <em>doing the work</em> and <em>knowing the work from a position of direct engagement</em>.</p><p><strong>Epistemic contact is not empathy.</strong> Empathy — the capacity to understand another’s emotional experience — is a relational quality. Epistemic contact is a structural condition. A leader may empathise with practitioners’ frustration without understanding the work that produces it. Epistemic contact requires understanding the work itself, not merely the experience of those who do it.</p><p><strong>Epistemic contact is not servant leadership.</strong> Servant leadership (Greenleaf, 1970) is a dispositional orientation — a commitment to serving those one leads. Epistemic contact is a structural requirement that applies regardless of disposition. A servant leader without epistemic contact is well-intentioned but epistemically ungrounded; a leader with epistemic contact but without servant leadership dispositions at least makes decisions informed by the reality of the work.</p><p><strong>Epistemic contact is not management by walking around.</strong> MBWA (Peters &amp; Waterman, 1982) encourages leaders to engage informally with people throughout the organisation. While this is preferable to isolation, it is occasional, typically performative, and shaped by the leader’s existing assumptions about what is important. Epistemic contact requires structured, consequential, sustained engagement — not casual observation.</p><h3>5. The Aspiration Gap: The Central Diagnostic Contribution</h3><h3>5.1 Definition</h3><p>The <strong>aspiration gap</strong> is defined as the structural condition in which an organisation’s career architecture defines advancement as progressive distance from practice, creating a systematic divergence between what the organisation needs (epistemic contact at every layer) and what its members are incentivised to pursue (escape from the conditions that produce epistemic contact).</p><p>The term “gap” is chosen deliberately. It denotes the measurable distance between two things that should be aligned: the direction of career advancement (away from practice) and the direction of epistemic contact maintenance (toward continued engagement with practice). As individuals advance, this gap widens — they move further from the work while occupying positions that require increasingly consequential judgments about it.</p><h3>5.2 The Mechanism</h3><p>The aspiration gap operates through five compounding mechanisms.</p><p><strong>Structural incentivisation.</strong> Organisations distribute rewards — compensation, status, autonomy, influence, and job security — in ways that increase with hierarchical altitude. The practitioner track, where it exists, typically offers lower ceilings across all of these dimensions. The implicit message, confirmed by observable organisational behaviour, is that directing work is more valuable than performing it.</p><p><strong>Aspirational contamination.</strong> Junior members observe the reward gradient and internalise its logic. They begin to view their current practice not as a capability to develop but as a credential to acquire — time served before the “real career” begins. This process, documented ethnographically by Jackall (1988) in corporate settings, progressively reshapes what organisational members want from their work, substituting advancement for deepening as the primary aspiration.</p><p><strong>Talent misallocation.</strong> The organisation’s most capable practitioners — those whose judgment has been most extensively developed through consequential engagement with the work — are systematically removed from practice through promotion. Their expertise is redeployed for coordination and direction, functions that require different capabilities and that the former practitioner may or may not possess. The Peter Principle (Peter &amp; Hull, 1969) captures the individual dimension of this dynamic; the aspiration gap captures its structural and cultural dimensions.</p><p><strong>Coordination inflation.</strong> Because advancement requires leaving practice, organisations accumulate coordination and oversight roles beyond what the work requires. Graeber (2018) documented the proliferation of roles whose primary function is to manage, interpret, and report on the work of others — what he called “bullshit jobs.” The aspiration gap produces a specific variety of this phenomenon: roles that exist because the career architecture demands them, not because the work does.</p><p><strong>Epistemic depletion at the top.</strong> As the aspiration gap moves the most capable practitioners into management and leadership, the upper layers become populated by people whose last direct experience of the work grows progressively more distant. Their judgment, once grounded in consequential engagement, is increasingly constructed from representations. The confidence-competence gap described in Section 3.3 widens, and the organisation’s capacity for reality-grounded decision-making at the strategic level degrades.</p><h3>5.3 The Aspiration Gap and the Practitioner Intelligence Framework</h3><p>The aspiration gap interacts destructively with the Consumer–Mechanic–Engineer–Scientist progression described in the Practitioner Intelligence framework (Bickersteth, 2025). This progression describes deepening <em>within</em> practice: the movement from instrumental tool-use (Consumer) through diagnostic capability (Mechanic) and design capability (Engineer) to generative theoretical contribution (Scientist). Each stage represents a more sophisticated relationship between the practitioner and the theoretical foundations of their domain.</p><p>The aspiration gap short-circuits this progression. Instead of advancing from Consumer through Mechanic to Engineer to Scientist — a trajectory that produces the deepest forms of practitioner intelligence — organisational career architectures route talented practitioners out of the progression entirely, directing them toward management tracks that move laterally rather than deepening vertically. The organisation retains competent practitioners at the Consumer and Mechanic levels but fails to produce the Engineers and Scientists whose deep engagement would advance the quality of practice itself.</p><p>The loss is not merely individual. When an organisation’s most capable practitioners are systematically diverted from the deepening trajectory, the organisation loses its capacity to generate the most sophisticated forms of Practitioner Intelligence — the kind that would drive genuine innovation, challenge embedded assumptions, and advance the field’s understanding of its own work.</p><h3>6. Extending the KWD Framework and Reflexive Competence to the Organisational Level</h3><h3>6.1 KWD Distribution Across Layers</h3><p>The Knowing–Working–Doing (KWD) framework (Bickersteth, 2025) distinguishes three modes of professional intelligence: Knowing (codified understanding), Working (applied judgment), and Doing (observable action under consequence). In the closed circuit, all three modes are integrated in a single person. In the layered organisation, each layer has a primary mode:</p><p>The practitioner’s centre of gravity is Doing. The manager’s centre of gravity is Working. The leader’s centre of gravity is Knowing. The owner typically operates external to the KWD framework, holding consequential and material stakes without primary engagement in any of the three modes.</p><p>Epistemic contact, specified through KWD, requires that each layer maintain sustained engagement with its non-primary modes. The practitioner needs contact with Knowing (strategic context) and Working (coordination logic). The manager needs contact with Doing (execution reality) and Knowing (strategic frame). The leader needs contact with Working (translation dynamics) and Doing (operational substance). The moment any layer operates exclusively in its primary mode, its judgment becomes ungrounded and the meaning it generates attenuates.</p><h3>6.2 Organisational Reflexive Competence</h3><p>Reflexive competence, as I defined it in the Practitioner Intelligence paper, is the individual practitioner’s capacity to maintain awareness of their own judgment without disrupting its fluid execution — resolving the tension between the tacit character of expert performance and the reflective awareness needed for continuous development.</p><p>I now extend this construct to the organisational level.</p><p><strong>Organisational reflexive competence</strong> is defined as the capacity of an organisation’s layers to maintain awareness of the work they mediate without collapsing the structural differentiation between layers or drifting into epistemic disconnection from the work.</p><p>This construct resolves an analogous tension at the organisational level. On one hand, each layer must maintain sufficient distance from the work to perform its distinctive function: the manager must coordinate without doing, the leader must direct without micromanaging. On the other hand, each layer must maintain sufficient proximity to keep its judgment calibrated by reality. Organisational reflexive competence names the condition in which both requirements are simultaneously met.</p><p>Several features distinguish organisational reflexive competence:</p><p>First, it is distinct from organisational learning as typically construed (Argyris &amp; Schön, 1978; Senge, 1990). Organisational learning concerns the acquisition and application of knowledge across the organisation. Organisational reflexive competence concerns the <em>maintenance of epistemic contact between layers</em> — a structural condition rather than a learning process.</p><p>Second, it is distinct from transparency or information sharing. Information can flow freely between layers without producing epistemic contact, because the information is representational rather than experiential. Organisational reflexive competence requires not just information flow but the maintenance of channels through which the textured, situated, tacit understanding of the work at the point of delivery remains accessible to upper layers.</p><p>Third, it is dynamic rather than static. The appropriate degree of proximity between layers varies with the stability of the operating environment. In stable conditions, greater distance may be tolerable. In volatile conditions, closer engagement is essential. Organisational reflexive competence includes the meta-capacity to calibrate this proximity — to recognise when epistemic drift is occurring and adjust the organisation’s mechanisms for maintaining contact accordingly.</p><h3>7. Implications</h3><h3>7.1 Career Architecture</h3><p>If the aspiration gap is a structural mechanism that degrades epistemic contact, career architecture reform is the primary intervention. Career systems informed by epistemic contact would provide genuinely equivalent practitioner tracks (with equivalent compensation, influence, and recognition), require mechanisms for maintaining epistemic contact across all career paths (including management and leadership tracks), and replace the default assumption that advancement means distance from practice with a design principle that advancement in any direction must preserve engagement with the KWD modes that the advancing role moves away from.</p><p>The Consumer–Mechanic–Engineer–Scientist progression provides the developmental model for practitioner tracks. The equivalent mechanism for management and leadership tracks is what I term <em>practice immersion</em>: structured, recurring, consequential engagement with the work at the point of delivery — designed not to develop practitioner skill in the manager or leader but to recalibrate their understanding of the work they mediate.</p><h3>7.2 Role Design</h3><p>Conventional role design specifies decision authority. Role design informed by epistemic contact would additionally specify <em>epistemic obligations</em>: what a person in the role is required to understand from direct engagement with the work, not merely from reports or representations. A manager’s epistemic obligation might include sustained, regular engagement with the execution reality of their team’s work. A leader’s epistemic obligation might include periodic, structured immersion in the operational conditions their decisions affect.</p><p>This reframes the role not only as a jurisdiction of responsible judgment (as proposed in the Practitioner Intelligence paper) but as a jurisdiction of <em>required understanding</em> — specifying not just what the person decides but what they must know from direct engagement in order to decide well.</p><h3>7.3 Learning Architecture</h3><p>Professional development informed by epistemic contact must include mechanisms for <em>perceptual recalibration</em> — the updating of managers’ and leaders’ understanding of the work to match its current reality. This is distinct from conventional management development, which focuses on skill acquisition (learning to manage) rather than perceptual maintenance (keeping one’s understanding of the work current). The distinction parallels the behavioural transformation / skill mastery distinction developed in the Practitioner Intelligence framework: managers and leaders need not behavioural transformation of how they manage but perceptual recalibration of what they understand about the work their management mediates.</p><h3>7.4 Organisational Evaluation</h3><p>Evaluation systems informed by epistemic contact would assess not only the quality of outputs but the quality of understanding maintained. For managers, this includes the accuracy of their model of the work — measurable through structured comparison between managers’ descriptions of the work and practitioners’ reports of it. For leaders, this includes the groundedness of their strategic decisions — assessable through retrospective analysis of whether leaders’ predictions about operational implications proved accurate.</p><h3>7.5 AI Strategy</h3><p>AI’s capacity to perform the informational and coordinative functions that have historically justified organisational layers renders the epistemic contact question urgent. When AI can summarise, coordinate, analyse, and generate strategic narratives, the only remaining justification for human roles at every layer is the exercise of situated judgment grounded in epistemic contact with the work. Organisations that design for epistemic contact will retain human roles whose judgment is grounded and irreplaceable. Organisations that do not will discover that their human layers, having been reduced to information processing, are vulnerable to automation — not because AI is more intelligent, but because the layers had ceased to exercise the situated judgment that would have made them irreplaceable.</p><p>The design principle that follows extends the principle from the Practitioner Intelligence paper: AI should be evaluated by its impact on epistemic contact across organisational layers, not merely by its effect on task completion or information processing efficiency.</p><h3>8. Conclusion: A Structural Claim Worth Testing</h3><p>I have argued that epistemic contact — a live, consequential, reality-disciplined relationship between organisational roles and the work they mediate — is a structural condition for organisational legitimacy and effectiveness, and that the aspiration gap — career architectures that define advancement as distance from practice — is the primary mechanism through which modern organisations erode this condition.</p><p>The paper’s central theoretical contributions are two. First, the concept of epistemic contact itself, which provides a formal construct for analysing and designing the relationship between organisational layers and the work they mediate — extending the Practitioner Intelligence framework from the recognition of individual expertise to the structural conditions required for sound organisational judgment. Second, the concept of the aspiration gap, which names and analyses the self-reinforcing cultural mechanism through which organisations teach their members to desire the very conditions that degrade judgment and erode meaning.</p><p>I advance two testable propositions:</p><p><strong>Proposition 1:</strong> Organisations in which managers and leaders maintain higher degrees of epistemic contact with the work (as measured by the accuracy of their models of operational reality) will exhibit higher-quality strategic decision-making, lower rates of implementation failure, and higher levels of meaningful work reported across all layers.</p><p><strong>Proposition 2:</strong> Organisations in which the aspiration gap is narrower (as measured by the equivalence of rewards, influence, and recognition between practitioner and management tracks) will retain deeper practitioner expertise, exhibit less coordination inflation, and produce more grounded strategic decisions.</p><p>Both propositions are empirically investigable through comparative organisational studies, longitudinal analyses of career architecture reforms, and structured assessments of epistemic accuracy across organisational layers.</p><p>In a world where AI can generate explanations, frameworks, and strategic narratives at scale, the organisational differentiator will not be who can produce the most sophisticated model. It will be whose models are grounded in the reality of the work — whose layers maintain the contact with practice that makes their judgment sound and their authority legitimate.</p><p>That is epistemic contact. It is what organisations need and what their current structures systematically erode. I have attempted to give the construct theoretical standing — and structural permission.</p><h3>References</h3><p>Arendt, H. (1958). <em>The Human Condition</em>. University of Chicago Press.</p><p>Argyris, C., &amp; Schön, D. A. (1978). <em>Organizational Learning: A Theory of Action Perspective</em>. Addison-Wesley.</p><p>Bickersteth, A. D. T. (2025). Practitioner Intelligence: Expertise, reflexive competence, and judgment in the age of AI. <em>Manuscript in preparation</em>.</p><p>Chandler, A. D. (1962). <em>Strategy and Structure: Chapters in the History of the American Industrial Enterprise</em>. MIT Press.</p><p>Crawford, M. B. (2009). <em>Shop Class as Soulcraft: An Inquiry into the Value of Work</em>. Penguin Press.</p><p>Csikszentmihalyi, M. (1990). <em>Flow: The Psychology of Optimal Experience</em>. Harper &amp; Row.</p><p>Dreyfus, H. L., &amp; Dreyfus, S. E. (1986). <em>Mind over Machine: The Power of Human Intuition and Expertise in the Era of the Computer</em>. Free Press.</p><p>Eraut, M. (2000). Non-formal learning and tacit knowledge in professional work. <em>British Journal of Educational Psychology</em>, 70(1), 113–136.</p><p>Fayol, H. (1916). <em>Administration industrielle et générale</em>. Dunod.</p><p>Fromm, E. (1955). <em>The Sane Society</em>. Rinehart &amp; Company.</p><p>Graeber, D. (2018). <em>Bullshit Jobs: A Theory</em>. Simon &amp; Schuster.</p><p>Greenleaf, R. K. (1970). <em>The Servant as Leader</em>. Robert K. Greenleaf Center.</p><p>Jackall, R. (1988). <em>Moral Mazes: The World of Corporate Managers</em>. Oxford University Press.</p><p>Klein, G. (1998). <em>Sources of Power: How People Make Decisions</em>. MIT Press.</p><p>Lave, J., &amp; Wenger, E. (1991). <em>Situated Learning: Legitimate Peripheral Participation</em>. Cambridge University Press.</p><p>Lips-Wiersma, M., &amp; Morris, L. (2009). Discriminating between ‘meaningful work’ and the ‘management of meaning.’ <em>Journal of Business Ethics</em>, 88(3), 491–511.</p><p>Marx, K. (1844/1959). <em>Economic and Philosophic Manuscripts of 1844</em>. Progress Publishers.</p><p>Mintzberg, H. (1973). <em>The Nature of Managerial Work</em>. Harper &amp; Row.</p><p>Mintzberg, H. (2004). <em>Managers Not MBAs: A Hard Look at the Soft Practice of Managing and Management Development</em>. Berrett-Koehler.</p><p>Muller, J. Z. (2018). <em>The Tyranny of Metrics</em>. Princeton University Press.</p><p>Nonaka, I., &amp; Takeuchi, H. (1995). <em>The Knowledge-Creating Company: How Japanese Companies Create the Dynamics of Innovation</em>. Oxford University Press.</p><p>Perrow, C. (1984). <em>Normal Accidents: Living with High-Risk Technologies</em>. Basic Books.</p><p>Peter, L. J., &amp; Hull, R. (1969). <em>The Peter Principle</em>. William Morrow and Company.</p><p>Peters, T. J., &amp; Waterman, R. H. (1982). <em>In Search of Excellence</em>. Harper &amp; Row.</p><p>Polanyi, M. (1966). <em>The Tacit Dimension</em>. University of Chicago Press.</p><p>Rosso, B. D., Dekas, K. H., &amp; Wrzesniewski, A. (2010). On the meaning of work: A theoretical integration and review. <em>Research in Organizational Behavior</em>, 30, 91–127.</p><p>Schön, D. A. (1983). <em>The Reflective Practitioner: How Professionals Think in Action</em>. Basic Books.</p><p>Senge, P. M. (1990). <em>The Fifth Discipline: The Art and Practice of the Learning Organization</em>. Doubleday.</p><p>Sennett, R. (1998). <em>The Corrosion of Character: The Personal Consequences of Work in the New Capitalism</em>. W. W. Norton.</p><p>Sennett, R. (2008). <em>The Craftsman</em>. Yale University Press.</p><p>Simon, H. A. (1947). <em>Administrative Behavior</em>. Macmillan.</p><p>Simon, H. A. (1956). Rational choice and the structure of the environment. <em>Psychological Review</em>, 63(2), 129–138.</p><p>Smith, A. (1776). <em>An Inquiry into the Nature and Causes of the Wealth of Nations</em>. W. Strahan and T. Cadell.</p><p>Taleb, N. N. (2012). <em>Antifragile: Things That Gain from Disorder</em>. Random House.</p><p>Vaughan, D. (1996). <em>The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA</em>. University of Chicago Press.</p><p>Weber, M. (1922/1978). <em>Economy and Society</em>. University of California Press.</p><p>Weick, K. E. (1993). The collapse of sensemaking in organizations: The Mann Gulch disaster. <em>Administrative Science Quarterly</em>, 38(4), 628–652.</p><p>Weick, K. E. (1995). <em>Sensemaking in Organizations</em>. Sage Publications.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a95bc8bcde70" width="1" height="1" alt=""><hr><p><a href="https://medium.com/human-at-work/epistemic-contact-meaning-judgment-and-the-structural-consequences-of-organisational-scale-a95bc8bcde70">Epistemic Contact: Meaning, Judgment, and the Structural Consequences of Organisational Scale</a> was originally published in <a href="https://medium.com/human-at-work">Human at work</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>