The expertise trap

The surgeons who resist robotic surgery the most aren’t the mediocre ones.

They’re the brilliant ones.

The ones who spent fifteen years developing hand precision that other surgeons describe with something close to reverence. The ones who can close a cranial incision so cleanly that residents gather just to watch. These are the surgeons who look at the da Vinci robotic system and feel, before they can articulate anything, a faint and genuine nausea.

The robot is more accurate than any human hand. The data on this is not ambiguous. And yet the resistance is real, and it is not stupid.

Because what the robot is taking isn’t just a task. It’s a life’s work. It’s twenty years of the thing that made you the person everyone calls when it matters most.

This is exactly what is happening to the best designers and product people right now.

Not the average ones. The exceptional ones.

The ones who have spent a decade developing a sense of interface so refined they can feel a bad interaction pattern before they can name it. The ones whose Figma files other designers screenshot and study in their free time. The ones who take real, earned pride in the quality of their craft.

These are the people who open an AI design tool, let it generate a few options, look at the output, and quietly close the tab.

The output wasn’t good enough. That part is often true. But it’s not the reason they closed the tab.

What you know lives in your body, not your brain.

Cognitive scientists call it procedural memory. The kind of knowledge that stops needing your conscious mind once it’s fully in you. Riding a bicycle. Touch-typing. The way you navigate your childhood home in the dark without turning on a light.

Design expertise is mostly procedural memory.

I realised this working with a junior designer on a consumer onboarding flow. She’d built something technically correct: right components, right hierarchy, spacing that matched the system’s rules exactly. But it felt wrong in a way I couldn’t immediately explain. I sat with it for ten minutes trying to find the language.

The problem was the visual weight on the second screen. Users’ eyes would scan left-to-right, drop down, scan again, and land on an element that created a tiny hesitation just before the main action button. A fraction of a second, invisible unless you’d trained yourself to feel for it. She couldn’t see it because she hadn’t yet built the internal library that would let her feel it. I couldn’t explain it quickly because the knowledge lived below language.

That’s expertise. And it’s genuinely extraordinary.

But when an AI generates twenty variations of that same flow in four minutes, the procedural knowledge that took ten years to build becomes one input among many. The expert’s advantage shifts from execution to selection. From making to judging.

Most designers I know haven’t made peace with that shift yet. Not because they’re obstinate. Because nobody warned them that their most valuable skill and their most dangerous blind spot were the same thing.

Designers should be solving for a layer that no longer needs them.

At a product company I spent time with a few years back, the senior designer was genuinely exceptional. The kind of person who notices a two-pixel misalignment in a file shared for developer handoff, which is the point in the process where a designer passes their work to the engineers who’ll build it, and fixes it before saying good morning. The team adored her. The quality of the product’s interface was, measurably, because of her.

When the company started using AI tools to generate initial design explorations, she did something interesting. She spent more time than ever on craft. Every screen she personally touched became more meticulous, more detailed, more precisely considered. Her output, on a per-screen basis, got measurably better.

Her output, on a per-week basis, got measurably smaller.

She was doing the equivalent of a surgeon insisting on hand-suturing every wound while the robotic system waited in the corner. The work was more beautiful. It was also slower, harder to scale, and increasingly invisible to the people evaluating what the team was actually shipping.

This is the expertise trap in its clearest form. The more you double down on the thing you’re brilliant at, the more you risk optimising for a problem the market has already started solving differently.

But the trap isn’t fear, exactly, and it isn’t laziness. It’s something more honest than that.

She loved the work. And the work was genuinely good.

The problem is that love of execution and love of outcome are not the same thing. They feel identical when you’re the best at the execution. They diverge sharply when execution is no longer the scarce resource.

The robot didn’t take the surgeon’s precision. It took the feeling of precision.

Something real is lost when surgical robots take over from hands. The surgeons who object aren’t entirely wrong about that.

What they’re describing, even if they can’t quite name it, is the physical experience of the work itself. The specific sensation of tissue resistance under a blade. The confidence of trained hands moving through a procedure they’ve performed five hundred times. That felt knowledge gets replaced by something more mediated, more distant, more like operating a very precise machine than doing surgery.

But what doesn’t get replaced is the judgment underneath it. When to operate and when not to. What a good outcome looks like for this patient specifically. What the scan is telling you that the numbers aren’t.

The instrument changed. The intelligence it required did not.

I think about this every time a designer tells me they’re worried AI will take their job. My honest answer, which I don’t always say directly, is this: it will take the part of your job you were always slightly too proud of.

The pixel manipulation. The component construction. The part that felt most like craft because it was most tactile, most visible, most satisfying in the moment of doing it.

But it won’t take your taste. It can’t.

Taste isn’t a skill you learn in a course. It’s the slow accumulation of every right and wrong call you’ve ever made or watched someone else make, compressed into judgment you can deploy without knowing you’re deploying it. You can’t generate it. You can only earn it, slowly, over years of caring about the wrong things until you figure out which things are actually right.

The designers who matter most in five years will be the ones who make taste the primary work, rather than the thing that happens between the frames.

Keeping your hands is fine. Making your hands the point is not.

Some people will read this as an argument to abandon craft entirely and pivot to AI prompt engineering. It isn’t.

A surgeon who spends Sunday afternoons on simulations because they love the feel of it, because it sharpens something in them that matters, is doing something worth preserving. Some things are worth doing for how they feel, not just what they produce.

But a surgeon who performs a procedure by hand when robotic assistance would measurably reduce patient risk, because they need to feel the instrument, has turned their love of craft into the patient’s problem.

That’s the line.

Early in my career, I spent months obsessing over interaction details that, I can say honestly now, nobody ever noticed. Micro-animations that would have required frame-by-frame analysis to appreciate. Hover states refined to a resolution that lived entirely below conscious perception. I was putting craft into things at a depth no user would ever reach.

I’m not ashamed of it. It taught me things I still use.

But I was also, if I’m being direct about it, building beautiful things for myself and calling it user-centred design. I have the Figma files to prove it. They’re gorgeous. They’re completely useless.

The expertise trap isn’t that you love what you do. The expertise trap is when what you love becomes the thing you’re protecting, even when the work would be better served by letting some of it go.

One designer uses AI and brings their taste to bear on what it generates. Another treats AI as a threat and builds more carefully by hand to prove a point. Both are making a claim about where value lives. Only one of them is right about where the work is going.

The hands were never the point.

When robotic surgery arrived, the question surgeons kept asking was: should we keep using our hands?

It was the wrong question.

The right one was: what were the hands always actually for?

Precision, most surgeons would have said. And they’d be right. But precision in service of what? Someone going home instead of to an ICU. A damaged system repaired. A person who walked in scared walking out with a plan.

Once you name the real purpose, the instrument question gets much simpler. You use whatever gets the patient home. Sometimes that’s your hands. Sometimes it’s a robot. Often it’s both, taking different roles across different moments in the same procedure.

Designers are circling a version of this right now, mostly without realising it. The argument in most product teams isn’t really about AI tools. It’s about where the value of design actually lives. Which is a question the industry has never had to answer precisely because execution and judgment used to be bundled together. The same person who had great taste was also the one who could build what the taste demanded.

AI unbundles them. And suddenly the question everyone avoided becomes the only one that matters.

What is the design actually for? Not what are you designing. What is the act of design, as a practice, ultimately meant to produce?

If the answer stops at “a well-crafted interface,” you’ve put the hands at the centre. And the hands, it turns out, were always the means, not the thing.

The neurosurgeons who adapted best didn’t abandon their hands.

They just stopped believing the hands were the point.

Precision was the instrument. The patient was always the point. Most of them knew this somewhere. They just hadn’t had to say it out loud until the robot arrived and made the distinction impossible to ignore.