The Tool Doesn't Know What It's Making

There's a version of this essay that opens with a stat. Something about how many images are generated every day, or how many lines of code AI now writes, or the percentage of designers who feel threatened by the shift. I'm not going to write that version. You've already read it. You already feel what those numbers point to. You're living inside the question those numbers raise.

So let me start somewhere more honest.

The uncomfortable truth at the center of this moment is not that AI is replacing designers and developers. It's that AI is revealing which parts of our work were always just execution, and which parts were actually thinking. For a lot of teams, that line is drawn in a more uncomfortable place than anyone wants to admit.

When the Floor Becomes the Ceiling

For years, the ability to make things well meant something. Not just as a skill, but as a signal. Knowing how to use the tools took time. Building fluency with Figma, or React, or a design system, or a particular visual language took years of practice, failure, and refinement. The quality of your output said something about you that a resume couldn't.

That signal just became noise.

Anyone with a prompt and an afternoon can now generate a landing page that looks professional. A pitch deck that looks polished. A component library, a brand identity, a mobile app flow, an API wrapper. The execution layer, the part where you turn an idea into a thing that exists, is no longer the hard part. It's no longer where value lives.

This should be liberating. And it is, a little. But mostly, if you're honest with yourself, it's disorienting. Because for many of us, the making was not just a means to an end. It was how we understood ourselves. It was the identity. It was the proof of our value. And now the proof is no longer sufficient.

The floor just became the ceiling. And teams that don't recognize this will spend the next few years optimizing for a game that has already changed.

What AI Cannot Do, and Why That's Harder Than It Sounds

Here's what AI cannot do, at least not yet, and not without you.

It cannot know why a product exists. It cannot hold the memory of eighteen months of user research and translate that into a judgment call about what to cut. It cannot feel the tension in a room when a feature ships that the team knew was wrong but shipped anyway. It cannot decide what's worth fighting for. It cannot ask the question that reframes the whole problem.

It cannot have taste in the way that matters most, which is not aesthetic taste but ethical taste. The taste to know when something is technically beautiful and still completely wrong for the people it's meant to serve.

It cannot care.

Now. Here's where I have to be rigorous with you, because this is where the argument gets soft in most essays. People list what AI cannot do and then exhale, relieved that their jobs are safe. But that relief is premature, because the question is not only whether AI can do these things. The question is whether your team is currently doing them either.

Are you asking why before you ask what? Are you spending time on the questions that don't have obvious answers? Are you bringing ethical rigor to every sprint, every feature decision, every dark pattern you noticed but let slide because the deadline was real?

If the answer is not a clear yes, then the gap AI is creating in your industry is also a mirror. And the thing looking back at you is not a robot. It's the version of your work you always meant to do but kept deferring to the execution.

The Prompt Jockey Problem

There's a real danger that teams respond to this moment by becoming very good at prompting and very little else.

Prompt engineering is a skill. I'm not dismissing it. Getting useful, specific, high-quality output from AI tools requires fluency, iteration, and judgment. But it's a thin skill if it floats on nothing. If you cannot evaluate the output, if you cannot tell whether a generated design actually solves the problem or just looks like it might, if you cannot articulate what's wrong with code that compiles but creates a bad experience, then you're not a designer or developer who uses AI. You're a dispatcher. And dispatchers are much easier to automate than thinkers.

The teams that will matter in five years are not the ones that generate the most, the fastest. They're the ones that can tell the difference. That can look at a hundred generated outputs and know which one is right, and know why, and know what to change about it, and know when to throw all of them away and go back to first principles.

That capacity has a quiet name: discernment. The ability to perceive meaningful differences and act on them. That's what needs to be cultivated right now, even as, or especially as, the generation becomes effortless.

What Adaptation Actually Looks Like

It doesn't look like adding AI to your existing process and calling it a day. That's lipstick. Real adaptation is structural, and it asks something harder of teams.

It asks you to move the conversation upstream. The most valuable design and development work now happens before a single pixel is placed or a line is written. It happens in the space of questions: What is this actually for? Who does it exclude? What does success look like for the person using it, not just for the business selling it? What are we not building, and why?

Those conversations used to get squeezed to the margins because the execution took so long. Now the execution is fast. Which means there's no excuse anymore for skipping the thinking. Teams that continue to skip it will produce more, faster, and it will still be the wrong thing, only at scale.

It asks you to invest in judgment as a practice. Judgment is not a talent you either have or don't have. It's built through exposure, reflection, and friction. Teams need to create conditions where people are regularly asked to defend their decisions, to explain the reasoning behind a choice, to articulate what they're optimizing for and what they're trading away. This slows things down in the short term. It makes the work better in the long term. And it builds the capacity to evaluate AI output in ways that actually matter.

It asks you to get clearer about values. When anyone can generate anything, the question of what should be generated becomes urgent in a way it never was before. AI will help you build manipulative onboarding flows just as readily as it will help you build accessible, honest ones. It doesn't care. It cannot care. That caring is entirely on you, and it needs to be legible, explicit, and enforced. Not a values statement on a wall. An actual reckoning, in the room, every time a decision is made.

It asks you to redefine what craft means. Craft used to live in the execution. The perfect kerning. The elegant function. The animation that felt just right. Those things still matter, but they're now table stakes, achievable by anyone with a decent prompt and a good model. Craft now lives in the decisions that precede all of that. The shaping of the brief. The framing of the problem. The editorial judgment about what to keep, what to change, what to refuse. Craft is curatorial now. And that demands a different kind of rigor, one that's less about hand and more about mind.

The Honest Invitation

Here's the thing I want to leave you with.

This moment is painful for a reason. It's painful because it's honest. It's stripping away the parts of our work that were comfortable but were not actually the point. The making was never the point. The point was always the effect the making had on people. The product that helped someone understand something. The interface that reduced friction in a moment of real need. The system that scaled access to something that used to require privilege. That was always the point.

AI is not taking that away from you. It's just making it impossible to avoid.

The teams that adapt will not be the ones that figure out how to use the tools most efficiently. They will be the ones that use the tools to finally clear space for the work they always meant to be doing. The slow, hard, irreplaceable work of deciding what matters and why, and then building toward it with intention.

That's the adaptation. Not faster generation. Deeper thinking.

The tool doesn't know what it's making. But you do. That's still worth something. Make sure it shows.