<?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 Erik Sagen on Medium]]></title>
        <description><![CDATA[Stories by Erik Sagen on Medium]]></description>
        <link>https://medium.com/@kartooner?source=rss-d55845b101c6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*yIY904WyirHpJ9dWgWB7zA.jpeg</url>
            <title>Stories by Erik Sagen on Medium</title>
            <link>https://medium.com/@kartooner?source=rss-d55845b101c6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 16:03:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kartooner/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[Don’t make me ask]]></title>
            <link>https://medium.com/paychex-ux/dont-make-me-ask-70de0c86c664?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/70de0c86c664</guid>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Fri, 16 Oct 2026 00:12:46 GMT</pubDate>
            <atom:updated>2026-10-16T00:12:46.741Z</atom:updated>
            <content:encoded><![CDATA[<h4>Proactive UX, crafted components, and closing the loop on ephemeral UI</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0v3jXcfcDFDo4_US47khHw@2x.jpeg" /><figcaption>Photo by Erik Brolin on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Whether you’re in design, development, or product, the easy move right now is to bolt a chat box onto everything.</p><p>Every app gets a little sparkle icon, or whatever your symbol is for intelligence, in the corner and a text field that says “Ask me anything.” It looks like progress, but mostly it just moves the work. Instead of clicking through menus, people now have to stop what they’re doing, open a panel, and figure out what to type. And that can be the hardest part.</p><p>Attaching chat to everything isn’t the solve. Intelligence built into the workflow is better, showing up inside the work people are already doing.</p><p>For a while now I’ve been writing about ephemeral UI and intent-first design. Screens don’t have to be fixed places you go. They can show up when you need them and get out of the way when you don’t. A lot of recent writing pushes that idea toward “just asking,” where software understands what you want and the switches fade into the background.</p><p>I agree with most of that. But asking is still work, and the part I’d push on is where I think the real design work is.</p><h3>Dashboards and haystacks</h3><p>Dashboards are good at what they do. They give power users every control and every number in one place. If you know the terrain, a good dashboard is fast, and nothing an agent says will beat it for someone who knows exactly which number they want.</p><p>But dashboards assume you know what you’re looking for. If something is wrong and you don’t know where, a dashboard hands you the haystack and wishes you luck on finding the needle.</p><p>That’s the gap agents can fill. Not by replacing the dashboard, but by doing the looking. An agent with good conditional logic and clear interaction rules can notice what’s off, figure out who needs to know, and bring it to them.</p><p>So it’s not dashboard or agent. The dashboard is for control and depth. The agent is for finding the thing you didn’t know to look for.</p><h3>Built from crafted pieces</h3><p>If the agent is going to bring things to people, it needs something to show them. This is where a lot of the “UI is dying” talk falls short. UI isn’t going away. What changes is who puts it together and when.</p><p>Video games figured this out a long time ago.</p><p>Games like Spelunky and Dead Cells have levels that feel different every time you play. But they aren’t made from nothing. Designers hand-build rooms, then write rules for how those rooms can connect. The game assembles a new level each run, and every room still works.</p><p>That’s the model for design systems when agents are doing the assembly. Designers craft the pieces: the components, the tokens, the states, and the rules for what can sit next to what. The agent puts the screen together in the moment, based on what the person needs. The layout isn’t fixed, but the quality is. Whatever gets built is still accessible, still on brand, and still usable.</p><p>That changes the job. Less drawing every screen. More defining the parts and the rules, then testing what happens when they combine in ways nobody planned. Level designers playtest their rooms over and over. We’ll need to do the same.</p><h3>Showing up where the work is</h3><p>The best version of this doesn’t force people into one app. People work in a lot of places. On a shop floor, on a phone between meetings, in an inbox at the end of the day.</p><p>The same intent can take a different shape depending on where it lands. A quick text for someone on their feet. A card in the app for someone at a desk. A summary in email for something that can wait. Deciding which channel fits is part of the design work now, and it’s a rule set too. When is something urgent enough for a text? When does it belong in a weekly digest?</p><p>The question stops being how we get people into our product. It becomes how our product shows up well wherever they already are.</p><h3>From reactive to proactive</h3><p>Most agent UX today waits to be asked. You type a question, it answers, maybe with a nice chart. That’s useful, but it’s still reactive. It’s a search box with better manners.</p><p>Proactive UX means the system notices first. A payroll example shows the difference. Say an employee forgets to clock out.</p><ol><li><strong>Reactive. </strong>An admin finds the gap while squaring up payroll and has to chase it down.</li><li><strong>Assisted.</strong> The system flags the missing punch and sends it to the manager to fix. Faster, but it’s still cleanup after the fact.</li><li><strong>Proactive.</strong> The system notices this person tends to miss their out-punch. Near the end of their shift, it sends a friendly heads up. The punch never goes missing.</li></ol><p>Same problem, three levels. Only the last one keeps the problem from happening. The admin never has to deal with it at all.</p><p>I use a simple framework for this: anticipate, prepare, surface. Lately I’ve added a fourth step, review.</p><ul><li>Anticipate is noticing the pattern. This employee misses out-punches on late shifts.</li><li>Prepare is getting the context ready. When does their shift usually end? Where do they actually see messages?</li><li>Surface is delivering it at the right moment, in the right place. A nudge on their phone a few minutes before the shift ends.</li><li>Review is where people stay in charge. Someone can see why the nudge fired, adjust the rule, or turn it off.</li></ul><p>That last step matters most. Without it, proactive turns into a system doing things to people instead of for them.</p><h3>What proactive costs</h3><p>Proactive is better. It’s also harder, and it costs trust when the rules are careless.</p><p>A few questions every team needs to consider and answer:</p><ul><li>How many misses make a pattern? Set it too low and it nags. Set it too high and you’re back to reactive.</li><li>How does it sound? “Heads up, you usually clock out around now” feels like help. “You’ve missed four punches this month” feels like a write-up.</li><li>Who sees the pattern? If the employee gets a nudge, does the manager see the history too? That’s where helpful can start to feel like being watched.</li><li>Who can change the rules? If nobody can tune it, nobody will trust it.</li></ul><p>These aren’t just engineering details. They’re the UX. Ten years ago we argued about hover states. Now we’re deciding how often a system gets to tap someone on the shoulder.</p><h3>Closing the loop</h3><p>Ephemeral UI was always half the idea. Screens that come and go are only useful if something smart decides when they should come and what they should show.</p><p>That’s the loop. Crafted pieces so the UI is always solid. Channel rules so it lands in the right place. Anticipate, prepare, surface, and review so it shows up at the right time, with people still in control when it does.</p><p>The choice was never portal or chat box. It’s knowing when to speak, what to show, and what to leave for a person to decide.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=70de0c86c664" width="1" height="1" alt=""><hr><p><a href="https://medium.com/paychex-ux/dont-make-me-ask-70de0c86c664">Don’t make me ask</a> was originally published in <a href="https://medium.com/paychex-ux">Paychex UX</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Closing the gap]]></title>
            <link>https://medium.com/@kartooner/closing-the-gap-dcb0a2422c7c?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/dcb0a2422c7c</guid>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Mon, 27 Jul 2026 22:54:00 GMT</pubDate>
            <atom:updated>2026-07-27T22:55:09.340Z</atom:updated>
            <content:encoded><![CDATA[<h4>What agentic tools are actually changing for designers who build</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*57Tp6EUZZySm13y_rm0Htw@2x.jpeg" /><figcaption>Photo by <a href="http://x1brett.redbubble.com/">Brett Jordan</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>For years there’s been a running debate about whether designers should learn to code. Just search for “Should designers learn how to code?” and you’ll find many articles and takes on the matter.</p><p>The usual answer landed somewhere in the middle: perhaps know enough to talk to developers, don’t try to be one. That answer was fine for its time. I think it’s worth revisiting now.</p><h4>Why fundamentals still matter</h4><p>Just as a furniture designer should understand wood grain, a designer who knows the tools that help create a thing is going to have a more enriching experience. If you understand how a browser actually renders your layout, how state changes occur, how interactions get built, you design differently. You stop drawing pictures of software and start working with the material it’s actually made of.</p><p>That hasn’t changed.</p><h4>What’s actually changed and maybe for the better</h4><p>For most of design history, a developer crossed that distance alone, working from a static comp and a set of notes, guessing at the parts that were unclear. How should this feel when it loads? What happens when an error shows up? What’s the timing on that transition? Those questions usually got answered later on, maybe mid-way through build, via a DM or comment in Figma.</p><p>AI tooling, across its permutations, whether that’s a custom in-house tool or something like VS Code plus GitHub Copilot, is closing that gap. Not by turning designers into developers, but by giving them a way to prove an idea holds up before it ever reaches one.</p><p>Here’s how I think about the three stages now:</p><ol><li><strong>Ideate.</strong> This part really hasn’t changed much. Sketchpad on hand, drawing in sand, Figma, Miro, Pen.dev, whatever tool you’re comfortable in. Loose, fast, exploratory. This is still where the real thinking happens, and it should stay that way.</li><li><strong>Prove it.</strong> This is the new-ish stage. AI tooling, including things like MCP-connected agents, can take an idea from that canvas and turn it into a real, working prototype. Real HTML, real CSS, real JS. Not a picture of an interaction, the interaction itself. You can feel the timing. You can see what happens on the edge cases you’d normally forget to mention.</li><li><strong>Build it.</strong> A developer takes what you made and finishes it. They can, because they understand the underlying code and the stack in a way you don’t. The stack is really just their tool set, the same way Figma is yours.</li></ol><p>That middle stage is the actual news here. It didn’t really exist before, at least not in the way it exists currently.</p><h4>Oh, this isn’t a mandate</h4><p>None of this means every designer should start coding, or that if you don’t, you’re behind.</p><p>That’s not the point.</p><p>Some designers will make this part of their daily practice. They’ll reach for a coded prototype the way they used to reach for a higher-fidelity comp, because it’s become second nature.</p><p>Other designers will never think in code and don’t need to. They’ll dip into this once in a while, for the one idea that’s worth it enough to need proof instead of a picture, then hand it off and go back to working the way they always have. Or they’ll pair up with someone who can help bring the idea to life. There’s value in that too.</p><p>The point was never to make code part of your identity as a designer. It’s that the option now exists, and you can use it in whatever amount actually fits how you think.</p><p>I used to be a designer/front-end developer myself, before I moved fully into design leadership. I would have loved having agentic tools like this to work with back then. For me, working this way beats piecing together components in Figma.</p><p>That’s my mileage, though, and it’ll vary for you.</p><p>Some people will find the same shortcut I did. Others will still get there faster working the way they always have, through Figma, or through sketches shared back and forth with an LLM. The point isn’t which path is correct. It’s that more paths exist now than there used to.</p><h4>A challenge, not a handoff</h4><p>I want to be clear about what the last stage is, because it’s easy to read this wrong.</p><p>The goal was never to hand a developer finished code and say “just ship this.” The prototype isn’t production-ready, and it shouldn’t be. That’s not sloppiness, that’s the point. The prototype’s job is to be honest about behavior and intent. Nothing more.</p><p>Handing it off is a way of saying: here’s exactly what I mean, now take it the rest of the way. It’s a challenge, and it’s meant as a form of respect. The developer understands the system, the architecture, the constraints across the whole product in a way I never will from inside a design tool. My job was never to replace that knowledge. It was to stop making them guess.</p><p>I get why a developer might read all this warily. Spoiler: you are not alone.</p><p>“Designer hands over code” has usually meant inheriting something that might need to be rebuilt, or defending scope against a PR nobody asked for. That’s a fair thing to be tired of, and it’s not what this is.</p><p>The prototype doesn’t touch the data layer or hook into a database. It doesn’t make decisions about authentication, about how errors get handled for real, about anything that has to hold up at scale. None of that is mine to decide, and the prototype should make that obvious just by what it leaves out.</p><p>What it’s allowed to say is narrower and more useful: here’s the timing, here’s what this should feel like, here’s the edge case I noticed while I was building it. You keep every call that actually matters. You just get to make it with more information than a static comp ever gave you.</p><h4>Why this matters</h4><p>It was less a fear and more a fit. Visual tools were, and often still are, the better tool for ideating. Sketching and moving shapes around lets you think in ways that writing code doesn’t. That’s still true, and it’s not what this changes. The canvas tools are still canvas tools, and you’ll still have reasons to reach for them. Code only enters once you’re past ideating and need to prove something holds up.</p><p>What’s actually new is that a designer can now hand over a coded prototype that says something true about how a thing should behave, built by someone who understands the fundamentals well enough to know what that prototype is even saying.</p><p>That’s not designers overstepping into engineering. That’s designers being more precise about intent, so the people who build the real thing can spend their time on what they’re actually good at.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dcb0a2422c7c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Business Acumen 101 for Designers]]></title>
            <link>https://medium.com/@kartooner/business-acumen-101-for-designers-00a4d3487ca2?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/00a4d3487ca2</guid>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[business-acumen]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Fri, 22 May 2026 18:32:05 GMT</pubDate>
            <atom:updated>2026-05-22T18:33:06.531Z</atom:updated>
            <content:encoded><![CDATA[<h4>How to push back without being the blocker</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yMJtQvPNKRlNLQNY6xXY4Q@2x.jpeg" /><figcaption>Photo by KaroGraphix Photography on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4>Imagine this: you’re in a product kick off meeting, all together reviewing an initiative that the team is sure will be valuable for the customer.</h4><p>Except, as you’re sitting in this meeting, with your partners in this work to be done, it feels off; the whole thing is unverified, messy, rushed. One of those or all three.</p><p>As a designer, even if your role description has leader or not, you decide to raise your hand and mention that proverbial elephant that’s standing there in front of you: “Do we have enough research for this?” Or even the big whopper: “Is the UX hitting the mark?”</p><p>When these things are said, the room will tune you out. To product and engineering leaders, often under intense pressure, you don’t sound like a strategic partner. You sound like a roadblock in the process of getting the work done.</p><p>If you want to change the outcome of these meetings, you just have to change your language.</p><h3>Speak with your partners, not to them</h3><p>We need to stop fighting business constraints and start using them as design guardrails.</p><p>That’s where business acumen comes in. At least a baseline understanding of what’s sometimes even called <strong>business savvy</strong>. It’s not about it memorizing spreadsheets or learning corporate jargon. Nah. Don’t waste your energy on that.</p><p>Rather, just understand that development, engineering and product focus are finite resources.</p><p>When we frame our questions around risks, resource allocation, value proposition, guarantee you that the business will listen.</p><p>Instead of just asking if a project is worth doing, we need to ask what must be true for it to succeed.</p><p>This changes things from passive order taking to active strategic partners. It allows us to push back without saying no, protecting our partner’s time and effort while also creating the space for divergent, lo-fidelity thinking.</p><p>Okay; enough of that, here’s a template to use:</p><h4>Is this project worth our time and money? How do we know?</h4><ul><li>For a sales deadline: Do we need a working system now, or will a polished mockup win the deal?</li><li>For a top down mandate: What specific metric or goal is leadership trying to move?</li><li>For one big customer: Is this meant just for them, or can we scale it to other clients later?</li><li>To cut internal costs: What is the current manual workaround costing us right now?</li><li>To copy a competitor: Are customers actually leaving us over this, or are we reacting out of fear?</li></ul><h4>What does success for this thing look like in 6 months?</h4><ul><li>Is success making the sale today, or ensuring customers use it long-term?</li><li>What concrete data proves this investment paid off?</li><li>What existing roadmap work are we putting on hold to make room for this?</li></ul><h4>What is the simplest thing we can test to prove value?</h4><ul><li>Can we test the core idea with a sketch, mockup, or manual process first?</li><li>What is the most expensive engineering part, and can we skip it for the first test?</li><li>What is the absolute smallest version we can ship to a tiny group of users? (This is where prototyping fits in)</li></ul><p>Next time you are in a messy, unclear kickoff meeting, drop the design jargon and lean more on the business savvy.</p><p>Use these questions to protect your team’s focus, preserve engineering time, and create the space you need to actually design the right solution.</p><p>And next time you’ll been seen as a strategic partner in the process of doing the good work.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=00a4d3487ca2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[It Was Never About the Screen]]></title>
            <link>https://medium.com/@kartooner/it-was-never-about-the-screen-743722d100b9?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/743722d100b9</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Tue, 05 May 2026 15:15:44 GMT</pubDate>
            <atom:updated>2026-05-05T15:25:15.318Z</atom:updated>
            <content:encoded><![CDATA[<h4>Designing the brain. Letting UI be the byproduct.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/790/1*gUs5zravxVli_FUiQNiVEg@2x.jpeg" /></figure><p>For most of the last couple of decades, the job of a designer looked increasingly like this: user clicks something, something happens. Cause meets effect. Then design the screen, ship the screen.</p><p>That production model survived for a long time. But it was never where the hard decisions lived; it was just where the decisions were easiest to see.</p><p>Understanding the user’s environment meant tweaking the user flow. Making a user flow change was reflected by increasing or decreasing the amount of screens. Changing the information architecture impacted the items in a navigation menu, or how the content was laid out on the page.</p><p>As <a href="https://www.linkedin.com/pulse/ai-changes-lot-things-how-what-we-design-doesnt-change-adam-connor-aaszc?utm_source=share&amp;utm_medium=member_ios&amp;utm_campaign=share_via">Adam Connor</a> put it:</p><blockquote>“The screens, flows, and components are the last mile of that work. Not the whole thing.”</blockquote><p>The work that shapes those screens, the decisions about what to surface, when, and why always lived somewhere else.</p><p>But now there’s a layer underneath the surface doing more complex, independent work. Pattern matching, flagging anomalies, making rules and calling on tools to help in that work. Deciding when to surface something to a human and when to not surface anything at all. That layer, the orchestrator, is where design needs to happen right now. And almost nobody is designing it on purpose.</p><p>The questions that feed it aren’t even technical. How much evidence does someone need before they’re asked to confirm something? What’s the threshold for the system to recognize a pattern and suggest a rule? When does the intelligence act and when does it wait?</p><p><strong>Those ARE the UX questions.</strong> They’re just unclaimed.​​​​​​​​​​​​​​​​ Most of the time delegated to whatever a chatbot generated, or an engineer making a stack of assumptions on a whim.</p><p>That can make or break the product experience.</p><p>The experience a person has is the system you’re designing. The separation between “the interface” and “what the system does” is morphing in new ways. They are intermingled. At times, it appears as the same thing, viewable from two sides.</p><p>User experience is the experience the user has. It’s not the drawing of the experience. The drawing is a byproduct of the cognitive design work; a way to clearly communicate thinking. But the thinking was always the work.</p><p>Designers got pulled toward artifacts because artifacts are visible. You can point to them, argue about them, feel comfortable by them because they seem real. You can show them in a portfolio. You can measure output in screens shipped.</p><p>And, to make matters worse, developers weren’t trusted to produce usable screens anymore so UX became the factory line to produce screen layouts that developers would code.</p><p>Meanwhile the harder, invisible work (understanding the user, mapping the journey, thinking about the system) got squeezed out or renamed as “product strategy” and moved upstream to somebody else so the illusion of user understanding could happen.</p><p>Then, the directives would be handed down. Brought back to you as an order taker. To make the screens.</p><p>But now, LLMs can produce artifacts fast. Real fast. Diagrams in seconds. Mockups in minutes. Prototypes in an hour.</p><p>The artifact got cheaper, the thinking and judgement didn’t. And, as the artifacts get cheaper, the thinking, the discernment to articulate what “good” means in the nebulous world of design becomes increasingly more valuable.</p><p>And the orchestration layer, the part nobody claimed, is where the judgement is critically needed.</p><blockquote>“Systems design becomes primary over tactical UI design.” — <a href="https://bigmedium.com/speaking/sentient-design-josh-clark-talk.html">Josh Clark, Sentient Design</a></blockquote><p>Designing the orchestration layer is where the foundational UX skills are critical. As you work to define the behavioral logic that will govern the users interactions with the system, you are re-applying structural design skills like interaction design and information architecture:</p><ul><li>When will the system propose an action, vs proactively take the action?</li><li>Who will it prioritize when coordinating between two different humans in the system</li><li>What library of elements will it use to display that information, and when can it invent something “new”?</li></ul><p>Instead of a static UI mockup with variants, you are defining guardrails and principles for these systems to curate (not dictate) what the experience can be.</p><p>Josh Clark again:</p><blockquote>“…the happy path gives way to happy possibilities.”</blockquote><p>This article was written in close collaboration with <a href="https://medium.com/u/a6b08612227f">Jon Daiello</a>, who provided some words and the awesome illustration. Jon is a Sr. UX Manager at Paychex and writes about UX and product design topics.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=743722d100b9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Empty Table]]></title>
            <link>https://medium.com/@kartooner/the-empty-table-a20a2f0a5d55?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/a20a2f0a5d55</guid>
            <category><![CDATA[design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[futures-thinking]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Thu, 30 Apr 2026 01:08:14 GMT</pubDate>
            <atom:updated>2026-04-30T01:25:46.151Z</atom:updated>
            <content:encoded><![CDATA[<h4>Knowing what’s good is not a feature AI can ship</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7QY05JQJDy1XG93KZGsUrQ@2x.jpeg" /><figcaption>Photo by <a href="http://twitter.com/benjaminchild">Benjamin Child</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Picture a conference room. Floor-to-ceiling glass. A long table running down the center. Chairs around it.</p><p>No people.</p><p>Just laptops. Open, running, humming along in some version of a meeting that no longer needs you.</p><p>That image lives somewhere in a lot of people right now. Designers. Writers. Strategists. Anyone who builds things with their brain. The fear isn’t some far-off idea. It has a shape. And that shape looks like a room you used to belong in.</p><p>But that image gets something wrong.</p><p>The room is still full of humans. You just can’t see them yet.</p><h3>It knows a lot. It doesn’t know what’s good.</h3><p>AI does one thing really well. It processes. It finds patterns across a massive amount of information and gives back what is most likely to be correct. Ask it a question and it returns the most probable answer based on everything it has ever seen.</p><p>That’s useful. It’s also not the whole job.</p><p>Likely and good are not the same thing.</p><p>A design decision. A creative direction. Knowing when something feels off before you can explain why. Those aren’t math problems. They’re judgment calls. And judgment is something you build over years. Getting things wrong. Figuring out why. Trying again. It lives in your point of view. In the way you read a room. In your gut.</p><p><strong>AI doesn’t have a gut.</strong></p><p>What it has is pattern recognition at a scale no person can match. So here’s the actual split: you bring the judgment. It brings the processing. Those are two different jobs.</p><h3>What happens when judgment leaves the room</h3><p>There’s another, quieter risk worth saying out loud. If humans stop making the call, things start to look the same. Design converges. Writing starts to sound familiar in a way that’s hard to place.</p><p>That begins now.</p><p>When everyone runs their work through the same engine, the edges get smoothed down. The unexpected gets averaged out. What made your work yours disappears into the middle.</p><p>Your point of view isn’t a liability in that world. It’s the only thing that pushes back against it.</p><p>And there’s something else; a lot of us got into this work because of other people. Yes, to earn a living. But also to sit in a room and figure something out together. To make something with another human being that neither of you could have made alone. A table full of laptops can’t give you that.</p><h3>What this actually looks like</h3><p>You think and AI helps you think faster, or from an angle you hadn’t considered. You make the call.</p><p>It handles the parts that are genuinely computational. Pulling structure from a mess of notes. Crunching. Summarizing. Drafting a first pass you can react to. You set the direction. You decide what good means for this project, this audience, this moment.</p><p>That’s not a consolation prize. That’s the work. It always has been.</p><p>The mistake is thinking your value was in the output itself. It wasn’t. It was in knowing which output was worth making. In caring whether it was right. And sometimes knowing it wasn’t.</p><h3>You belong in the room</h3><p>Go back to that conference room from earlier. The empty one.</p><p>Now put yourself in it.</p><p>Not to defend your chair. Not to prove something. Because no laptop running in that room knows what the room actually needs. It doesn’t know your client’s history. It doesn’t know what got shot down last quarter and why. It doesn’t know when the energy in the room changes and the whole direction needs to change.</p><p><strong>You do.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a20a2f0a5d55" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Drawings Were Never The Job]]></title>
            <link>https://medium.com/@kartooner/the-drawings-were-never-the-job-aa2ce6963320?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa2ce6963320</guid>
            <category><![CDATA[design]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Sat, 18 Apr 2026 21:27:28 GMT</pubDate>
            <atom:updated>2026-04-18T22:24:25.928Z</atom:updated>
            <content:encoded><![CDATA[<h4>What twenty-plus years in design actually taught me about this moment.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L94V08tMm9fhoA47AMVxdQ@2x.jpeg" /><figcaption>Photo by Yukon Haughton on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I’ve been doing design work for (checks notes, ahem) over 20 years now, and for the first time, the tools are actually keeping up.</p><p>There’s real anxiety around AI taking design jobs. Google Stitch or Claude Design doing what used to take days, weeks, months. I get why it reads as a threat. But I don’t see a replacement. I see the end of the assembly line.</p><p>Most of my career, a huge chunk of the work was just production. Moving things around. Getting it just right. Now that part is being automated away, and honestly, good.</p><p>If you’re just starting out, that’s a legitimate concern. The repetitive work, the craft, the production of assets — that was how you built intuition. Spacing, hierarchy, flow. You learned the principles of design and the tenets of usability by grinding through them. That homework still matters. It just doesn’t come built into the job anymore. You have to go looking for it.</p><p>End of day, the job was never really about the artifacts. People you worked with probably thought it was. Called UX what they thought it was. <strong>The drawings</strong>. Meanwhile the actual work was synthesis — taking a mess of business goals and human needs and figuring out which path actually matters. That’s the messy part. The head-to-table days. Tossing papers. Throwing tables. That doesn’t automate. Neither does bad Bezier work, but that one I’m fine letting go.</p><h3>So where does that leave me?</h3><p>I’m not racing the machine on speed, and that would be silly anyway. I need sleep. Machines don’t. What I’m doing is using what I’ve built up over twenty years to direct the intent and hold the line on quality. Less making, more shaping. Less pixel-pushing, more of the intuition that took a lifetime to develop.</p><p>Took long enough.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa2ce6963320" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Future of UX: Smart UI, Not No UI]]></title>
            <link>https://medium.com/@kartooner/the-future-of-ux-smart-ui-not-no-ui-0367e22d72a9?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/0367e22d72a9</guid>
            <category><![CDATA[future]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[user-interface]]></category>
            <category><![CDATA[ux]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Thu, 09 Apr 2026 21:08:32 GMT</pubDate>
            <atom:updated>2026-04-09T21:08:32.391Z</atom:updated>
            <content:encoded><![CDATA[<h4>I’ve been reading a lot about how the user interface is going away. In some ways it will.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JsOuhVu0JQpG28fm7VrVWg@2x.jpeg" /><figcaption>Photo by <a href="https://www.amaltama.com/">Daria Nepriakhina 🇺🇦</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>The idea is simple: talk to an AI system, it does the work in the background. No buttons, no menus. But I think that’s only half the story.</p><h3>You need to see the data</h3><p>For simple stuff, like changing a password, a chat box works great. You ask, the AI does it.</p><p>But for complex work, like managing payroll or taxes for a company, it’s different. You’re not just giving orders. You’re trying to understand what’s actually happening.</p><p>There’s something about how we solve problems. We use our eyes and our memory to spot patterns. You can’t verify a thousand paychecks by listening to a bot say “it’s done.” You need to see the numbers. You need to scroll through and find the one that looks wrong.</p><p>You can’t feel the weight of data through text. You need to see it.</p><p>And you need to double-check. To be sure before you hit submit.</p><h3>The UI shows up when you need it</h3><p>The real thing isn’t choosing between UI and no-UI. It’s making the UI smart about when it appears.</p><p>Today, you have to hunt for tools. You go to a website, click through five menus, and finally find what you need or you need to login through a portal and hunt and peck to finish a task. That’s a lot of work before you can actually work.</p><p>The future is different. The tool shows up right when you need it.</p><p>Let’s say you’re asking an AI about a budget problem. Instead of just typing back a number, it shows you a slider or a table right there. You change the numbers, see how it affects everything, and click submit. Then it disappears.</p><p>This is what an orchestration layer does. It watches all your systems; payroll, taxes, forecasts, alerts. When something needs your attention, it surfaces it. An anomaly. A recommendation. Something that needs approval. It gives you a simple way to act on it. You make the call, close it, and go on with your day.</p><p>Meanwhile, the work keeps happening in the background. Small pieces of UI pop up when they need to. But you have a clean home to come back to. A place that doesn’t scream for your attention. A place where the noise is already filtered out.</p><p>The UI only shows up when you actually need to make a decision.</p><h3>Why leaders need to see</h3><p>Most leaders I talk to want to “trust but verify.” They like the idea of AI handling the work. But they don’t want to close their eyes and hope it got the taxes right.</p><p>That makes sense. That’s the pilot checking the instruments before takeoff.</p><p>If you take away the interface, you take away the human’s ability to see what’s happening. The full picture isn’t a blank screen. It’s like a cockpit. The plane can fly itself, but the pilot still needs the dials and the screens. You need to see what’s happening.</p><p><strong>The UI is there so you stay in control.</strong></p><h3>What’s next</h3><p>The future of UX isn’t about getting rid of the interface. It’s about an interface that knows when to show up, and when to get out of the way.</p><p>Visibility wins. Always.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0367e22d72a9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Do You Actually Need AI?]]></title>
            <link>https://medium.com/@kartooner/do-you-actually-need-ai-f45ae9e19ccf?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/f45ae9e19ccf</guid>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[product-design]]></category>
            <category><![CDATA[critical-thinking]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Mon, 06 Apr 2026 22:45:49 GMT</pubDate>
            <atom:updated>2026-04-06T22:45:49.338Z</atom:updated>
            <content:encoded><![CDATA[<h4>Every team I work with is adding AI somewhere.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z6WJTtKle59LkckIvPOF0w@2x.jpeg" /><figcaption>Photo by <a href="https://www.ericart.nl/">Eric Krull</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I’m doing it in my own projects. It feels like the default move right now; if you’re not building something with AI, you’re falling behind.</p><p>There’s real pressure here. Your leadership wants to show investors you’re “AI-forward.” Your competitors are announcing AI features. Your team hears about startups getting funded because of AI. It all adds up to one message: you need to move fast and add AI somewhere, anywhere.</p><p>AI features are enticing. They’re even marketable. They’re sellable. There’s nothing wrong with wanting to build them. But there’s a difference between building something because it’s marketable and building something because it actually helps your customers. You don’t want to market and sell something that’s going to damage trust.</p><p>You ship a feature that’s 70% accurate. Users come back confused. They stop trusting your recommendations. You don’t recover from that quickly.</p><p>Just because you can use AI doesn’t mean you should.</p><p>I’ve seen products get worse because teams bolted on AI features nobody actually needed (hat tip to my local grocery store).</p><p>The features felt impressive at first, but then users started asking, “Why is it suggesting this? I don’t trust it.” And suddenly you’ve got a feature that confuses people instead of helping them and they’ll avoid it.</p><p>The real problem is that we’re starting with the tool instead of starting with the problem. We see AI and think, “How can we use this?” when we should be asking, “Do we actually need this?”</p><p>Most of the time the answer is no. But saying no feels risky right now.</p><h3>Five questions to ask first</h3><p>Before you even start building an AI feature, ask yourself these five questions. If you can’t answer “yes” to all of them, you probably don’t need AI.</p><p><em>These questions exist for a reason: they keep you from shipping something that breaks trust.</em></p><h4>1. Does it actually help?</h4><p>Will this feature make someone’s job easier? Or is it just cool? There’s a big difference. A feature that’s impressive in a demo but doesn’t solve a real problem is just clutter.</p><p><strong>Be real honest here.</strong> If you’re building it because it’s trendy or because your leadership wants to say you’re “AI-forward,” that’s not a good enough reason.</p><h4>2. Does it need to be perfect?</h4><p>Some tasks need to be right 100% of the time.</p><p>Medical advice. Financial decisions. Legal documents. If a mistake could cause real harm, AI probably shouldn’t be making the call alone.</p><p>AI is getting better, but it still makes mistakes. If the stakes are high, a human should be in charge.</p><h4>3. Is there a clear path to the answer?</h4><p>AI works best when there’s a logical way to get from A to B. It takes information, follows a pattern, and produces an output.</p><p>But some problems don’t work that way. If you’re asking the AI to guess or make intuitive leaps in its logic, stop. If there’s no clear reasoning behind the answer, you shouldn’t trust it.</p><h4>4. What happens when it breaks?</h4><p>Every system fails sometimes. You’ve likely seen the blue screen of death (BSOD) at airports when Windows kicks out of kiosk mode.</p><p>So ask yourself: What’s the worst thing that could happen if the AI gets this wrong?</p><p>If it’s a suggestion for a design layout, that’s fine, the user can ignore it. But if it’s a recommendation that costs money or causes confusion, you need a backup plan. Usually that means keeping a human in the loop.</p><h4>5. Does it have the right information?</h4><p>AI can only work with what you give it. If the system doesn’t have access to the right data, it’s just guessing.</p><p>For example, if you’re building an AI writing assistant but it doesn’t know your brand voice or your company’s guidelines, it’ll probably give generic suggestions. It doesn’t know any better. Make sure the system knows what it needs to know.</p><h3>You can’t control every detail</h3><p>Here’s the straight dope: you can’t tell an AI exactly how to behave and expect it to follow your instructions word for word. It uses patterns to find answers. That’s the core of what it does.</p><p>If you’re trying to predict exactly what the AI will output or script every detail, you’re fighting against the technology itself. That’s usually a sign you don’t actually need AI. You probably need a rule-based system that does the same thing every time.</p><p>AI is useful because it can find patterns and adapt. But if you need that predictability and control, don’t force AI into the job.</p><h4>What happens when you get it wrong</h4><p>Customer trust doesn’t come back quickly. Once you break it, the damage spreads.</p><p>When you ship an AI feature that doesn’t work well. AI that’s confusing, or unreliable, or wastes people’s time, you don’t just lose that one feature. You lose confidence in your whole product.</p><p>Users start asking: If this AI feature is broken, what else is wrong here? Can I trust this product? Maybe I should look for something else.</p><p>I’ve read stories about teams spending months building an AI assistant that was right 70% of the time. Users stopped using it immediately. They didn’t trust it. Now the team has technical debt, lost credibility with their users, and a feature nobody uses. And that damage extends beyond that one feature. It makes users question what else might be broken.</p><p>The cost isn’t just the wasted development time. It’s the swift erosion of trust. And you don’t get that back easily.</p><p>This is where the five gates matter. They’re your defense against shipping something that damages trust.</p><h4>Show your homework</h4><p>If a feature passes those five questions, your job isn’t done. The biggest mistake teams make next is hiding how the AI got to its answer.</p><p>When an AI system suggests something, it should explain why. Show the user what information it based the suggestion on. Link to the sources it used. Add some helper text explaining the thinking.</p><p>Think of it like a math test. You don’t just write down the answer. You show the steps. Users need to see the steps too, so they can understand if the suggestion makes sense.</p><p>This also changes how users interact with the feature. Instead of accepting what the AI says without thinking, they’re actually reviewing it. They become the arbiter, not the AI.</p><h4>Be honest about what you don’t know</h4><p>AI systems aren’t always confident in what they suggest. Sometimes the system is 60% sure about a recommendation. Sometimes it’s 90% sure. Again, it depends on the data it’s able to look at. The UI should reflect that.</p><p>Instead of presenting the AI’s output like it’s the final answer, be honest about the uncertainty.</p><p>Use language like:</p><ul><li>“Based on your notes, I drafted this version. But I’m not totally confident about the color choices.</li><li>“I found three design patterns that might work here. The first one seems like the best fit, but you should double-check it against your guidelines.”</li><li>“I’m suggesting this heading structure, but there’s a chance it won’t work for your specific use case.”</li></ul><p>The user’s job (and what should be the default) becomes “review and edit” instead of “trust the machine.”</p><h4>Keep the human in charge</h4><p>The goal isn’t to remove humans from the process. It’s to build tools that actually make people’s lives easier.</p><p>By being clear about how the AI works, by showing your reasoning, and by being honest about uncertainty, you keep the human in control. The AI becomes a tool that supports decision-making, not something that replaces it.</p><p>If you can’t explain why the AI did something, you probably shouldn’t have let it do it in the first place.</p><h4>How to push back (without sounding afraid of technology)</h4><p>You might be reading this thinking: “This all makes sense, but my team is already building an AI feature. Or my leadership is pushing for one. How do I actually push back?”</p><p>It’s a real tension point. Saying “we don’t need AI” in a room full of people who believe the opposite takes courage. But you’re not being against technology. You’re protecting your customers. And protecting your customers is how you protect your bottom line.</p><p>When you use the five gates, you have a framework. You have language. You can say: “I’m all in on AI where it actually helps. But this feature doesn’t pass the five gates. Let’s look at question two: does it need to be perfect? If yes, AI isn’t the answer yet. If no, then question four: what happens when it breaks? The cost is higher than the benefit.”</p><p>This isn’t about being against AI. It’s about being realistic about when it works.</p><p>You can also be explicit about the risk: “If we ship something that confuses users or breaks trust, we’re not going to recover that quickly. Let’s make sure this is worth it.”</p><p>Most people understand that logic. Leadership included. They’ll respect the reasoning.</p><p>If they don’t? If the pressure to ship AI is so high that you can’t push back? That’s a different problem. But usually, the conversation changes when you make the risk clear.</p><p>So, before you add AI to your product ask yourself these five questions. Be honest about the answers. And if you can’t say yes to all of them, don’t do it.</p><p>The best features (AI or not) are the ones that solve real problems for real people. Don’t use AI just because it’s there. Use it because it’s the right answer.</p><p>If it’s not, don’t use it. That’s how you protect your product.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f45ae9e19ccf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Contribution, Not Compliance]]></title>
            <link>https://medium.com/@kartooner/contribution-not-compliance-d62d20a69a03?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/d62d20a69a03</guid>
            <category><![CDATA[partnerships]]></category>
            <category><![CDATA[trust]]></category>
            <category><![CDATA[parenting]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Wed, 25 Mar 2026 00:22:52 GMT</pubDate>
            <atom:updated>2026-03-25T00:25:28.570Z</atom:updated>
            <content:encoded><![CDATA[<h4>What happens when you stop controlling the method</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aXRLFd17jaZSx3hCUVLskQ@2x.jpeg" /><figcaption>Photo by Wai Siew on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><blockquote>“Leadership, very simply, is about two things: 1. Truth and trust. 2. Ceaselessly seeking the former, relentlessly building the latter.” — Jack Welch, former CEO of GE</blockquote><h4>I used to think the right way to get things done was to show people exactly how to do them.</h4><p>Chores had a method; certain steps, in the correct order. If you followed the steps, the outcome would be right. <strong>Right?</strong></p><p>For a while, it worked. The house stayed clean. The dishes got washed. Everything happened the way I planned.</p><p>But I was the only one who cared. The people doing the work, usually my kids, weren’t really there. They were going through the motions. Following instructions, not really contributing. Just complying. There’s a big difference between compliance and belonging, and I didn’t notice it until I stopped doing it that way.</p><h3>The truth</h3><p>One day I realized I was managing a household like I was managing a production line. Do this, then do that. Here’s the right way. The expected outcome didn’t change. We still needed a clean house, dishes still needed washing, laundry still needed doing. But I stopped telling people how to make it happen.</p><p>Instead, I said something different: “We all live here. We all benefit from existing under this roof. Here’s what needs to happen; we contribute. You figure out how.”</p><p>The dishes still get washed. The laundry still gets done. But now it happens differently than I would do it. And that’s the point.</p><p>Some people load the dishwasher a way I wouldn’t. Some people fold clothes differently. Some people have ways of doing things I never would have thought of. They surprise me even. It all works. The house functions. Things get clean.</p><p>But more than that, people own it now. They’re not following a script. They’re solving the problem themselves. That’s actual contribution.</p><h3>Why matters. How doesn’t.</h3><p>I realized the problem was that I was controlling the outcome and the method. Do it this way, get this result. It works. The dishes are clean. But here’s what I didn’t see: I was also controlling whether anyone cared.</p><p>So I stopped.</p><p>Instead of saying “wash the dishes like this,” I said something different: “The dishes need to be clean. We all live here. It matters. You figure out how.” And then I stepped back.</p><p>That’s an agreement, not a mandate. A mandate says: do this my way, because I said so. It demands obedience. It works in the short term. It also doesn’t build anything, not long term anyways. An agreement says: here’s why this matters. Here’s what needs to happen. I trust you to figure out how.</p><p>The <strong>why</strong> has to be real. Not “because I’m the parent” or “because I’m the boss.” The actual why. We live together, in this structure, on this flying rock. The space we share needs to function so we can all live well in it. You’re part of that. That’s not a punishment or a chore list. That’s reality.</p><p>And here’s the hard part for me: once I stop controlling the how, I don’t get to decide what that looks like. Someone loads the dishwasher differently. Someone else has their own system. I have to be okay with that. I have to trust that the outcome matters more than the method.</p><p>But the person doing the work gets something different too. They’re not following a script anymore. They’re solving the problem themselves. They own it. That’s contribution.</p><h3>This isn’t just about home</h3><p>The same pattern shows up everywhere. In workplaces. In communities. Anywhere people are trying to build something together.</p><p>Walk into a workplace where a manager tells people exactly how to do their job. Every step. Every method. No room for approach. People show up, do the work, and leave. They’re not building anything. They’re executing instructions and you have to ask yourself: do they feel valued at the end of the day?</p><p>Walk into a workplace where a manager says, “Here’s the problem we’re solving. Here’s why it matters. I trust you to figure out your approach.” People start to think differently. They own the work. They come up with better solutions because they actually <strong>care</strong> about the outcome.</p><p>A neighborhood cleanup where everyone’s assigned a section and told how to do it feels like a chore. A neighborhood where someone says, “Our street matters. Let’s take care of it. Do what makes sense to you and here’s a rake,” feels like belonging.</p><p>The difference is the same. Control the method and you get compliance. Clarify the why and step out of the way; you get participation. One person manages. The other builds culture.</p><h3>The hard part</h3><p>This is slower. Messier. Again, someone might do something differently than you would. You have to be okay with that.</p><p>And here’s what makes it hard: you’re giving up certainty. When you mandate the method, you know what the outcome will be. When you trust someone else to figure it out, you don’t. You’re betting that they’ll care enough to get it right. That’s a different kind of risk.</p><p>It requires patience. It requires letting go. It requires trusting that the outcome matters more than the method. And it requires something harder: believing that people actually want to contribute. Not because you’re forcing them. But because they understand why it matters.</p><p>That’s the real work. Not the chores or the tasks at hand. The letting go of control. The waiting. The trusting in people to show up.</p><h3>What actually happens</h3><p>When you do this, when you’re clear about the why and flexible about the how, something changes. People aren’t waiting to leave. They’re choosing to stay. They’re thinking about the problem. They’re taking ownership.</p><p>And then they do it with other people too.</p><p>A kid grows up in a house where contribution is expected but the method is theirs. They become someone who knows how to show up without needing to be told exactly how. They do that in their own home someday. Their own team. Their own community.</p><p>A team member works for someone who trusts them to solve problems their own way. They start leading that way too. They carry it forward.</p><p>A community member sees someone rally people around something that matters, then trust them to figure it out. They start doing that in their own circles and that propels forward.</p><p>It spreads. Not because someone mandated it. Because it works. Because people feel the difference between being managed and being trusted.</p><h3>The real thing</h3><p>The goal isn’t compliance, it’s contribution. And the only way to get contribution is to actually invite it. To be clear about what matters. To make space for how people want to show up.</p><p>And then to step back even when it’s uncomfortable.</p><p>Your child sees you do this. Your team sees you do this. Your neighbor sees you do this. And then they do it too. With their own kids. Their own teams. Their own circles. They pass forward the idea that people are worth trusting, that there’s more than one way to do the right thing.</p><p>That’s how culture actually shifts. Not from top-down rules about how to behave. From people who modeled it, trusted others, and created space for something better.</p><p>It’s not the easiest way. But it’s the only way that builds something that lasts.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d62d20a69a03" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When It Comes to AI, There Is No Playbook]]></title>
            <link>https://medium.com/@kartooner/when-it-comes-to-ai-there-is-no-playbook-c8a3d0d650ca?source=rss-d55845b101c6------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8a3d0d650ca</guid>
            <category><![CDATA[creativity]]></category>
            <category><![CDATA[ux-design]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Erik Sagen]]></dc:creator>
            <pubDate>Sat, 14 Mar 2026 03:11:19 GMT</pubDate>
            <atom:updated>2026-03-14T03:13:49.999Z</atom:updated>
            <content:encoded><![CDATA[<h4>Radio, television, refrigerators and now AI. We’ve been here before.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eYx1JfkQohnJcpyFvhYu4w@2x.jpeg" /><figcaption>Photo by <a href="http://frankokay.com/">Frank Okay</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Everyone’s talking about AI like they know exactly what it’s going to do. Spoiler: They don’t.</h3><p>You’ve read it here and in other places; that designers will be supercharged, developers will move faster, workflows will multiply. The future is certain, knowable, already here. Roll out the carpet.</p><p>But if you actually work somewhere trying to figure out how to use this stuff, you know something different. What’s actually bubbling up under the surface.</p><p>You know that nobody really knows what they’re doing yet. Your team is experimenting. Other teams are experimenting. The whole world is experimenting, and we’re all pretending we have a master plan.</p><p>Here’s what I think we’re missing in the conversation: <strong>this is normal</strong>. This is how new technology actually works.</p><p>We’ve just forgotten.</p><h3>The pattern that keeps repeating</h3><p>Go back to the early days of radio. People had transmitters and receivers, but they didn’t immediately know what radio was for. News? Entertainment? How would it make money?</p><p><strong>Nobody had a playbook.</strong> Certainly nothing you could pull off the shelf at a local bookstore.</p><p>Early broadcasters were just trying things, seeing what stuck, learning as they went. Some experiments worked. Most didn’t. Eventually some patterns emerged, and suddenly everyone knew what a radio station was supposed to be.</p><p>Then came television.</p><p>Same story, different technology.</p><p>Engineers had built something remarkable, but what was it actually for? How would people use it in their homes? What would they watch? Again, nobody knew. The infrastructure was there. The technology worked. But the application, the actual way it would integrate into people’s lives, was completely uncertain.</p><p>Lucille Ball and Desi Arnaz give us a perfect example of this uncertainty in action. Ball wanted to perform in front of a live audience the way she did on stage, which felt natural. But she also wanted the show to be filmed well enough to distribute and rewatch (later on we called this syndication). These two things seemed impossible to do at the same time. So they experimented. They tried filming with multiple cameras. They developed new techniques on the fly, solving problems nobody had solved before, because there was no instruction manual for what they were trying to do.</p><p>They weren’t following a playbook. They were making it up. And what they figured out became the template that television production followed for decades. The template everyone eventually copied came from two people in a room saying, “I don’t know if this will work, but let’s try.”</p><p>Jump forward to something smaller: the refrigerator replacing the icebox.</p><p>You’d think that would be instant; obviously a refrigerator is better, right? But it wasn’t. People had to learn new ways to cook. Shopping habits changed. How you preserved food changed. The way families organized their kitchens changed.</p><p>Adoption took time, because technology isn’t just a machine you install. It’s a disruption to how you actually live and work. People had to figure out what it meant for them.</p><p>Every transformative technology follows this pattern. The tools arrive. People start experimenting. There’s a period of real uncertainty where nobody knows what’s going to work. Then gradually, through all that trial and error, patterns emerge. Standards develop. Best practices show up and are documented by someone. And suddenly it feels like everyone always knew what they were doing.</p><p>But they didn’t. They were just making it up and time tells a different story.</p><h3>Where we are right now (and will be for the foreseeable future)</h3><p>We’re in that making-it-up phase with AI.</p><p>Right now.</p><p>But here’s the strange part: everyone’s talking like we’re not. The conversation is all confidence. AI will transform your workflows. AI will supercharge your team. AI will multiply your productivity. These things might be true eventually, but we don’t actually know yet. We’re still in the experimentation phase, and we’re treating it like we’re already at the part where the patterns are locked in.</p><p>Inside actual organizations, it’s messier. Teams are trying things. Some work. Some don’t. People are figuring out what an AI workflow actually looks like. Who owns it? How do you measure if it’s working? How long before you see actual return on what you’re spending? What happens when you ship something quickly and then need to improve it later?</p><p>No one knows. If you do, please contact me.</p><p>These aren’t rhetorical questions. These are the real problems teams are solving right now, in real time, without a playbook.</p><p>You can see this in the tools and frameworks that keep emerging. LangGraph. Multi-agent architectures. New ways of orchestrating AI systems. These aren’t appearing because the problem is solved. They’re appearing because we’re still trying to figure out what works. If we knew the answer, we wouldn’t need new tools every few months. The fact that the technical landscape keeps shifting is evidence that we’re still in the trial-and-error phase.</p><p>And that costs money. It takes time. Timelines stretch. Teams spend resources exploring paths that don’t pan out. That’s real, and we shouldn’t pretend it isn’t. This isn’t free. There’s a genuine cost to not knowing yet.</p><p>But that’s also how you figure things out.</p><h3>The strength in uncertainty</h3><p>Here’s what gets lost in the hype: uncertainty isn’t a weakness. It’s actually a strength, if you’re willing to work with it.</p><p>Uncertainty forces honesty. When you don’t know if something’s working, you have to keep paying attention. You have to keep asking whether this is actually better or if you’re just convincing yourself. You can’t declare victory and move on. You have to iterate. You have to improve.</p><p>There’s a difference between two approaches here. One is shipping a rough draft and abandoning it because you’ve “solved” the AI problem. You declare success, you move on, and later you discover that rough draft created a mess you now have to fix.</p><p>The other is shipping what works now and committing to making it better.</p><p>You stay engaged.</p><p>You keep improving.</p><p>You learn.</p><p>That’s iterative, that’s acceptable.</p><p>The second approach takes longer. It doesn’t really line up with your original timeline. But it actually works.</p><p>The confidence you see in headlines and keynotes will eventually show up in organizations. Real patterns will emerge. Best practices will form(ish). Teams will figure out what works for them, and then other teams will learn from that. It’s already starting to happen. But it won’t happen because someone declared it into existence. It’ll happen because people did the patient, unglamorous work of actually figuring it out.</p><p>That work is happening right now. In your company. In the startup down the street. In teams that are just trying to solve their actual problems and discovering things nobody’s tried before. They’re the ones having the Lucy and Desi moment; not because they’re trying to pioneer something grand, but because they’re just trying to make something work. Square peg meets star-shaped hole.</p><p>That’s how innovation actually happens. Not through certainty. <em>Through uncertainty.</em> Through people willing to experiment without knowing the endpoint, and having the patience to improve what they find.</p><h3>Small moves</h3><p>We’re in early days. This needs time. Not because we’re waiting for someone to hand us the answer, but because the answer only emerges through doing. Try some stuff. Some of it won’t work.</p><p>The teams and people that figure this out won’t be the ones pretending they already have it figured out. They’ll be the ones staying curious. The ones shipping thoughtfully. The ones committed to making things better over time, even when it takes longer than expected.</p><p>Don’t rush to certainty. Don’t pretend you know what you don’t know. The patterns and ways of working will emerge, but they’ll emerge from the actual work of figuring things out, not from declarations about what’s supposed to happen.</p><p>We’ve been here before.</p><p>And every time, the people who got to the proverbial there were the ones willing to make it up as they went.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8a3d0d650ca" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>