<?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 Karen Manning on Medium]]></title>
        <description><![CDATA[Stories by Karen Manning on Medium]]></description>
        <link>https://medium.com/@karen.pandkm?source=rss-422866f16831------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*egbyK9FCqdCPTugosjyFig.png</url>
            <title>Stories by Karen Manning on Medium</title>
            <link>https://medium.com/@karen.pandkm?source=rss-422866f16831------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 22:05:33 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@karen.pandkm/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[The AI Maturity Model I Use to Diagnose Any Organization]]></title>
            <link>https://medium.com/@karen.pandkm/the-ai-maturity-model-i-use-to-diagnose-any-organization-bd9c82488170?source=rss-422866f16831------2</link>
            <guid isPermaLink="false">https://medium.com/p/bd9c82488170</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[enterprise-architecture]]></category>
            <category><![CDATA[digitial-transformation]]></category>
            <dc:creator><![CDATA[Karen Manning]]></dc:creator>
            <pubDate>Wed, 26 Aug 2026 04:54:35 GMT</pubDate>
            <atom:updated>2026-08-26T06:02:48.193Z</atom:updated>
            <content:encoded><![CDATA[<p><em>“Are we behind on AI?” is the wrong question — and it leads to expensive answers. Here’s the five-dimension model I use instead, plus a free tool that identifies your binding constraint and the concrete next move to unlock it.</em></p><p>Every executive team asks me some version of the same question: <em>“Are we behind on AI?”</em> It feels like the responsible thing to ask. It’s also almost useless — because it produces an answer like “yes, probably,” which leads to an action like “let’s buy a bigger AI platform,” which is frequently the exact wrong move.</p><p><em>(I’ve distilled this into a free, no-login diagnostic tool — score the five dimensions in under five minutes. More at the end.)</em></p><p>The better question is: <strong>“Which dimension is our binding constraint, and what does the next level cost?”</strong> That one produces a fundable, specific next move. To answer it, you need a model that scores maturity per <em>dimension</em> — because the average score hides the constraint, and the constraint is the only thing that matters.</p><h3>Five dimensions, five levels</h3><p>The framework assesses organizations across five dimensions — each rated Level 1 (ad hoc) to Level 5 (transformative):</p><ol><li><strong>Data &amp; Infrastructure</strong> — can we feed and run AI reliably?</li><li><strong>Talent &amp; Skills</strong> — do we have the people, and is the org getting smarter?</li><li><strong>Governance &amp; Risk</strong> — can we deploy AI safely and defensibly?</li><li><strong>Adoption &amp; Process</strong> — does AI actually change how work gets done?</li><li><strong>Strategy &amp; Funding</strong> — is there a managed, funded plan, or a wish?</li></ol><p>The levels follow a consistent shape, adapted from CMM/CMMI maturity model structure: <strong>Ad hoc → Emerging → Operational → Systemic → Transformative.</strong> Level 2 is pilots with no repeatability. Level 3 is repeatable, measured delivery with some governance. Level 4 is a portfolio that’s managed, governed, and funded. In my experience across enterprise AI diagnostics, most organizations cluster around Level 2 overall — and, crucially, they’re <em>uneven</em>: a Level-3 data org stuck at Level-1 governance is completely typical.</p><p>Unlike composite-score maturity models that average across dimensions, this approach treats each as an independent variable — because the average hides the constraint, and the constraint is the only number that matters.</p><p>That unevenness is the entire point. Here’s why.</p><h3>The average lies. The shape tells the truth.</h3><p>Consider an illustrative example — call it <strong>Northwind Financial</strong>, a fictional regional bank convinced it’s behind and tempted to fund a large model-platform buildout. We score it:</p><pre>Data &amp; Infra      ████████░░  4<br>Talent &amp; Skills   ██████░░░░  3<br>Governance        ████░░░░░░  2   ◀ binding constraint<br>Adoption          ████░░░░░░  2   ◀ second constraint<br>Strategy/Funding  ██████░░░░  3</pre><p>Average: 2.8. If you lead with that number, you learn nothing. Look at the <em>shape</em> and the diagnosis is obvious.</p><p>Northwind’s instinct — buy more infrastructure — would pour money into <strong>Data &amp; Infrastructure, which is already their strongest dimension (4).</strong> That’s the classic mistake the average hides.</p><p>Their actual binding constraint is <strong>Governance (2)</strong> — and in a regulated bank, governance is high-leverage: until there’s risk-tiered intake and oversight, nothing can safely reach production, which is <em>exactly why Adoption is also stuck at 2.</em> Strong data feeding ungoverned pilots that never ship is the whole problem, in one sentence.</p><h3>The binding-constraint rule</h3><p>So the recommendation isn’t “improve everything.” It’s surgical:</p><blockquote><em>Raise your weakest </em>high-leverage<em> dimension by one level. Stop over-investing in dimensions already ahead of the constraint.</em></blockquote><p>For Northwind: stand up governance (risk-tiered intake, a written AI policy, model documentation — the building blocks of NIST AI RMF’s Govern function; see the <a href="https://github.com/kmanning-projects/ai-strat-responsible-ai-governance">companion governance playbook</a>) to move Governance 2 → 3 — which unblocks production — and shepherd two or three already-built pilots through it to move Adoption 2 → 3. Explicitly <em>do not</em> fund the platform expansion this cycle; Data/Infra is already ahead of the constraint.</p><p>Same budget. Completely different outcome. One path buys infrastructure they can’t yet safely use. The other converts existing, already-funded data strength into shipped value. That’s what a maturity model is <em>for</em> — not benchmarking your ego against competitors, but finding the one investment that unlocks everything else.</p><p>In my experience, moving Governance from Level 2 to Level 3 is a focused program quarter of work: a risk-tiered intake process, a written AI policy, and model documentation for in-flight systems — a smaller investment than the infrastructure expansion it replaces, and the investment that actually unlocks production.</p><h3>Why “next level” beats “best practice”</h3><p>Maturity models often fail because they hand you a Level-5 picture and an implicit instruction to transform into an AI-native tech company. That’s not actionable. The useful move is always <strong>one level up on the constraint.</strong></p><p>If your Governance is at Level 2 (ad hoc reviews, informal rules), Level 3 is concrete: a risk-tiered intake, a written policy, model documentation. That’s a quarter of focused work — in practice — not a five-year transformation. You raise the constraint, the next-weakest dimension becomes the new constraint, and you repeat. Maturity is a sequence of single steps on the binding constraint.</p><h3>Try it on your own org</h3><p>A free, no-login web tool runs this assessment: score the five dimensions with sliders, and it draws your radar profile, identifies your binding constraint (lowest score, broken by leverage), and tells you what the next level concretely looks like for that dimension. It also flags the dimensions you’re <em>over-investing</em> in relative to your constraint — the spending most orgs never question.</p><p>It’s a single static page; nothing leaves your browser. Try it here: <a href="https://km-ai-maturity-assessment.netlify.app/"><strong>km-ai-maturity-assessment.netlify.app</strong></a>. The full rubric (all five dimensions × five levels, with evidence signals) and the Northwind worked example are open-source alongside it.</p><h3>The takeaway for leaders</h3><p>Stop asking “are we behind?” It produces a platform purchase. Start asking “which dimension is our binding constraint, and what does the next level cost?” It produces a focused program that unlocks the value you’ve already paid for.</p><p>The pattern repeats across every engagement: most organizations are over-invested in models and infrastructure while stuck at Level 2 on governance and adoption — the two dimensions that actually convert capability into shipped value.</p><p><em>Interactive tool: km-ai-maturity-assessment.netlify.app — no login, nothing leaves your browser. Full rubric, worked example, and source: github.com/kmanning-projects/ai-strat-ai-maturity-assessment.</em></p><p><em>If you’re a CIO or Chief AI Officer who recognizes the “strong data, stuck governance” pattern — I’d like to understand what’s blocking the next level for your organization.</em></p><h3>Sources and attributions</h3><ol><li><em>Five-dimension AI maturity model:</em> Original framework developed by the author.</li><li><em>Level 1–5 progression (Ad hoc → Emerging → Operational → Systemic → Transformative):</em>Author’s original dimension definitions and scoring rubric, adapted from CMM/CMMI maturity model structure. Carnegie Mellon University Software Engineering Institute.</li><li><em>Northwind Financial:</em> Fictional organization used for illustrative purposes only.</li><li><em>“Most organizations cluster around Level 2 overall”:</em> Author’s observation. Not a published benchmark.</li><li><em>“One focused program quarter” for Governance Level 2→3:</em> Author’s estimate based on governance program design experience.</li><li><em>NIST AI RMF Govern function:</em> National Institute of Standards and Technology. (2023). <em>Artificial Intelligence Risk Management Framework (AI RMF 1.0)</em>. NIST AI 100–1. https://doi.org/10.6028/NIST.AI.100-1</li></ol><p><em>📊 AI Strategy Portfolio · AI Maturity<br>Part of a series. Open-source frameworks and companion articles: github.com/kmanning-projects/ai-strat-portfolio</em></p><p><em>*Note: This article is part of my technical thought leadership portfolio, synthesizing industry standards and original diagnostic frameworks into practical leadership guidance.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bd9c82488170" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Three Questions Every Enterprise AI Program Needs to Answer]]></title>
            <link>https://medium.com/@karen.pandkm/the-three-questions-every-enterprise-ai-program-needs-to-answer-c5b9a59c4cb3?source=rss-422866f16831------2</link>
            <guid isPermaLink="false">https://medium.com/p/c5b9a59c4cb3</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology-strategy]]></category>
            <category><![CDATA[digital-transformation]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[Karen Manning]]></dc:creator>
            <pubDate>Tue, 11 Aug 2026 06:01:31 GMT</pubDate>
            <atom:updated>2026-08-11T06:01:31.544Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Three questions determine whether an AI program creates durable value or becomes expensive technical debt. Here’s the framework I use to answer all three.</em></p><p>The conversation I have most often with enterprise AI leaders doesn’t start with “how do we build better AI?” It starts with a quieter version of the same question: “We’ve been at this for three years. Why don’t we have more to show for it?”</p><p>The honest answer is usually the same: the program never properly answered three foundational questions. These aren’t technology questions. They’re architectural and organizational — the kind that get deferred because they’re harder than picking a model provider or standing up a vector database.</p><h4>Where are you, across the dimensions that actually matter?</h4><p>“Are we behind on AI?” is the wrong question. It feels responsible. It produces expensive answers — the kind that lead to buying a bigger AI platform when governance is the actual constraint.</p><p>The better question: which dimension is the binding constraint, and what does the next level cost?</p><p>The framework scores organizations across five dimensions — Data &amp; Infrastructure, Talent &amp; Skills, Governance &amp; Risk, Adoption &amp; Process, Strategy &amp; Funding — each Level 1 (ad hoc) to Level 5 (transformative). The composite score is almost useless. The shape of the profile is everything: a Level-4 data organization stuck at Level-1 governance is the most expensive pattern I see, and the most common. Every infrastructure dollar spent while governance is the binding constraint is a dollar that can’t ship value.</p><p>The maturity model, the diagnostic, and the self-assessment tool are the first article in this series.</p><h4>What should you fund, in what sequence?</h4><p>Maturity tells you where the constraint is. The operating model tells you how to fund moving across it. This is where most AI programs fail at the portfolio level: everything is prioritized, nothing is sequenced, and the program is a collection of pilots without a production trajectory.</p><p>The AI Strategy Operating Model is a five-stage framework that takes an AI ambition statement and produces a sequenced, stage-gated roadmap — with use-case intake forms, a prioritization scorecard that separates quick wins from money pits, and a 90-day plan that ties each funded tranche to a proven outcome.</p><p>The critical insight is that strategy is a sequencing problem. The question isn’t “should we invest in AI?” It’s: given our maturity profile and our constraint, what’s the highest-leverage investment next quarter, and how do we fund it in a way that builds durable capability? Without sequencing, AI investment compounds as debt; with it, each funded move unlocks the return on the last one.</p><h4>How do you govern it — safely, defensibly, at scale?</h4><p>This is the question that derails the most programs, and the one most misunderstood. Governance is typically framed as risk management: review boards, approval gates, compliance documents. That framing produces governance that runs at committee speed — months behind the systems it’s supposed to oversee.</p><p>My position: governance is a design problem. The right design encodes it into the delivery system — so it runs at code-generation speed, without slowing anything down.</p><p>A commit-hook agent that checks architectural conformance before code merges is governance. An intake process that risk-tiers AI use cases before a line of code is written is governance. A model card that captures what the model does, what data trained it, and what risks it carries is governance. These are engineering artifacts. The question is whether they’re ambient — built into the process — or bolted on as an afterthought.</p><p>The governance work in this series covers the full stack: the responsible AI playbook mapped to NIST AI RMF, EU AI Act, and ISO 42001; the transformation of the architecture review function when AI is generating the code it’s supposed to govern; and the delivery system that closes the gap between a governance policy and engineering teams actually following it.</p><h4>How this work is different</h4><p>Most AI thought leadership lives in one of two places: executive strategy from people who’ve never shipped a model, or technical depth from people who’ve never made a business case. The work here operates in the gap. The frameworks published so far are backed by technical artifacts I built — the maturity assessment tool is live and free; governance and cost-benchmark proof points are in development and will be published here as they’re completed.</p><p>In my observation, the organizations succeeding with AI aren’t the ones with the best strategy documents. They’re the ones with the tightest loop between what the strategy says and what gets built. I write from inside that loop.</p><h4>What this body of work covers</h4><p><strong>Maturity and diagnosis:</strong> Where you are and what’s blocking you — five dimensions, a free self-assessment tool, the binding-constraint diagnosis.</p><p><strong>Portfolio and operating model:</strong> What to fund and in what sequence — a roadmap framework, prioritization scorecard, and 90-day plan templates.</p><p><strong>Governance as toolchain:</strong> How to deploy AI safely at scale — governance encoded into the delivery system, mapped to NIST AI RMF and the EU AI Act, extended to the agentic SDLC.</p><p>If you’re a CIO, Chief AI Officer, or enterprise architect who recognizes the pattern — serious investment, disappointing production output — this is where I work through why that happens and what to do about it.</p><p><em>The frameworks here are open-source and free — the goal is to make serious AI strategy accessible, not to gate it. If you’re working through any of these questions in your organization, I’d like to understand what’s blocking you.</em></p><p><em>🔢 AI Strategy Portfolio · The Three Questions<br>Part of a series. Open-source frameworks and companion articles: </em><a href="https://github.com/kmanning-projects/ai-strat-portfolio"><em>github.com/kmanning-projects/ai-strat-portfolio</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c5b9a59c4cb3" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>