The Accident That Looks Like Expertise

I keep thinking about a part of my chat with Justin Ranton for my podcast that I didn’t expect to stick with me. He builds clinical AI at SmarterDX, and he told me about the folks who check its results all day. With time, they get really good—they develop a sense for which alerts matter and which are just noise, a kind of calm, quick judgment that feels like deep knowledge. But Justin said some of what they’ve picked up isn’t really medicine; it’s the system itself—how this specific tool shows information, its quirks and habits. They’ve become experts at reading the tool, and they’re right often enough that it looks like clinical skill, until the system shifts and that feel quietly fades.

I’ve been mulling this over because I don’t think it’s just about AI. I think it’s about expertise, and it puts into words something I’ve felt about my own skills more than once.

We tend to picture skill as something inside us—portable, earned, ours. Some of it is. But a lot of what we call expertise is really a fit: a hard-earned balance between a person and a system they’ve lived in for years. The expert knows the landscape so well they stop noticing they’re reading it. From the inside, real, shareable judgment and a fine-tuned match to one tool feel exactly the same. That’s the catch. While you’re doing something well, it’s hard to tell how much is you and how much is the shape of what you’re good at.

You only find out when the shape changes.

When tools change, teams shift, or designs get a makeover, even the most seasoned pros can suddenly feel lost—and no one quite knows why. Their skills were real, but they were propped up by a hidden setup. Take that setup away, and you see how much it was holding up. That’s what I call brittle expertise: not a lack of talent, but talent tied to something unspoken.

Great tools can actually make this trickier, which is something designers might want to think about. The smoother a tool feels, the more it quietly handles behind the scenes. Before long, people become experts at the toolitself rather than the original task. We often worry that automation makes us lazy or just rubber-stampers. This is more subtle: the expert keeps thinking, but their focus shifts—becoming fluent in the tool’s world and less so in the real one. They’re still working hard; they’re just narrowing their expertise without realizing it.

The anthropologist in me finds this almost beautiful in a nervous sort of way. Humans are amazing at adapting—reading a system and shaping ourselves to fit it. That’s one of our best traits. But it means expertise is partly about the environment: it lives in the connection between a person and their tools, not just inside the person. We like to think skill belongs to the individual because it feels good and looks great on a résumé. Some of it travels, but some was always tied to the room where it grew.

I’ve switched careers enough to see both sides. I learned marketing on billboards, print ads, flyers, and even a NASCAR sponsorship—media that grab your attention and don’t let you click away. When digital came along, I figured my know-how would carry over, and some did: the old rule about how many times someone needs to see a brand before it sticks still worked. But a lot of what I called “skill” was really just instinct shaped by a world where attention was guaranteed. In a cluttered space where people can leave with one click, those habits didn’t just fall short—they misled me, promising attention the new medium no longer gave. It took time to sort out what I could keep from what I had to let go. Every fresh start shows you which strengths are truly yours and which were just borrowed from a setting you hadn’t noticed.

None of this means tools are bad, and I try not to romanticize the pre-tool expert. The same tools that create fragile expertise often make us way more capable than we’d be on our own. I wouldn’t want anyone going back to reading charts by hand. The message isn’t “don’t build great tools.” It’s that being an “expert” is less steady and less portable than we think—and some of the trust we give experienced users is really trust in a connection that could break without warning.

I don’t have a quick fix here—just a hunch and a better question. You can’t smooth out fragility just by polishing the surface; a cleaner screen won’t solve it. If there’s a real answer, it’s further back: in what a tool expects from us and what it quietly handles for us, like whether it keeps us skilled in the real work or slowly swaps that skill for skill in the tool itself. That’s a design choice, and it’s what I explore in my companion piece at Baleen Advisory. [link to “Designed to Be Doubted”] Justin and I dive into the technical side on the podcast.

But the short version is simpler—mostly a reminder to myself. Next time I see someone being effortlessly great at something a machine helped them do, I’m trying to ask a softer question before I get impressed: great at what, exactly—and what happens to them when the machine changes its mind?

Justin’s Healthcare UX episode.