How to find the industry that already solved your problem
You are redesigning the handoff between two teams inside your product, the moment a task's ownership passes from one queue to another, and every version you sketch still drops something in the middle. You have looked at every competitor in your category. They all drop the same thing, in the same place, because they are all solving it with the same vocabulary you are.
Why the fix isn't in your own field
A 2026 paper on AI-generated interfaces put a name to a pattern anyone building software this year has noticed: design homogenization, the tendency of AI-assisted design work to converge on the same look. The mechanism behind it is not really about AI. It happens whenever everyone solving a problem stays inside the same industry's reference set, whether that reference set is a component library or a Dribbble feed. Convergence is what a search space looks like when nobody leaves it.
The way out has a name too, and it is older than any AI design tool. Cognitive psychologist Dedre Gentner published the account of it in 1983, calling it structure-mapping theory. Her finding: a strong analogy is not two things that look alike. It is two things where the relations between the parts are alike, even when the parts themselves are unrelated. A hospital operating theatre and a Formula 1 pit lane share nothing visually. They can still share an identical relational shape, if the underlying coordination problem is the same.
The bias that keeps you looking too close to home
Here is the uncomfortable part. Cognitive scientist Kevin Dunbar did not test this in a lab with students solving puzzles. He spent roughly a year embedded inside four working molecular biology laboratories, recording how real scientists actually reasoned day to day. Analogy showed up constantly, and it was one of the strongest predictors of which scientists made real progress.
But of the 99 analogies Dunbar recorded, only 2 reached outside biology entirely. The default human move is to search one field over, not across the map. That is not a personal failing. It is what makes distant analogies rare enough to be worth deliberately going looking for one, instead of waiting for a stray thought to hand you one for free.
What Great Ormond Street borrowed from a pit lane
The clearest documented case of someone doing that deliberately is medical, not commercial. In the early 2000s, cardiac surgeons Martin Elliott and Allan Goldman at Great Ormond Street Hospital were watching handovers from surgery to intensive care fail in small, recurring ways: no single person owning the transition, information passed verbally and incompletely, equipment moved without a fixed sequence.
Stated as a hospital problem, that is a hospital's problem to solve alone. Restated functionally, as transferring full responsibility for a fragile, time-critical system between two expert teams with zero tolerance for dropped information, it stopped being one. The surgeons visited Ferrari's Formula 1 team, then McLaren's, to study pit-stop coordination directly.
They did not import equipment. They imported structure: one named leader for the handover, a fixed checklist, rehearsal, and someone whose only job was stepping back to check the whole picture rather than one task. A 2007 study in Pediatric Anaesthesia measured the result: technical and information errors during handover fell by roughly two-thirds.
Distance is not the part that matters
It is tempting to read that story as "the weirder the source industry, the better the idea." A 2010 study in R&D Management, examining 25 documented cross-industry innovation cases, checked that assumption directly and found no correlation between how cognitively distant a borrowed idea was and how valuable the resulting innovation turned out to be.
| What people assume | What the 25-case study found |
|---|---|
| A more exotic source industry produces a better result | No correlation between distance and outcome |
| The goal is to search as far away as possible | The goal is a genuine match in relational structure |
| Surface novelty signals a good analogy | Surface novelty is the thing that most often misleads |
A near-domain analogy with a real structural match will usually beat a distant one that only resembles your problem on the surface. Distance is not free value. It is just where you had to look, some of the time, to find a real match.
The field building your AI tools already runs on this
If the method sounds abstract, it is worth noticing that the tools sitting on your desktop are built from it. David Hubel and Torsten Wiesel's 1959 study of a cat's visual cortex found neurons organised in a hierarchy, simple cells detecting edges, complex cells pooling them into position-independent patterns. That biological finding fed directly into the architecture that became the convolutional neural network, the thing that lets a model recognise an object in a photo regardless of where it sits in the frame.
Diffusion models, the technique behind most current image generators, came from the same move in a different direction. Jascha Sohl-Dickstein and colleagues' 2015 paper took a description from nonequilibrium thermodynamics, how ink diffuses reversibly from a complex blob into a simple, even spread, and applied it to images: destroy an image into noise step by step, then train a network to run that process backward.
Neither of those started as an AI idea. Both started as someone rephrasing a solved problem from an unrelated field and asking whether the same shape applied. TRIZ, the engineering methodology built by analysing hundreds of thousands of patents, does the same thing on purpose, and a 2024 paper on AutoTRIZ shows a language model automating the first step, spotting the underlying contradiction and retrieving the inventive principle that has already resolved it elsewhere.
The search, run in twenty minutes
You do not need a patent database to start. You need to climb one rung above your own vocabulary before you start generating solutions.
- Write the problem exactly as you'd state it internally. Five minutes. Jargon included. This is the surface version, not the fix.
- Ask "why" until you hit a function a stranger would recognise. Two or three rungs, usually. "Fix the handoff modal" becomes "transfer full ownership of a task between two people under time pressure without losing information." That sentence has nothing product-specific left in it. Five minutes.
- Search that sentence, not your product category. Try AskNature.org if the problem is about passive regulation, efficiency, or structure under constraint. Try Untools' abstraction laddering guide if you want the ladder itself as a working tool. Ten minutes.
- Before you adopt anything, check the relations, not the resemblance. Ask whether the mechanism that made the source solution work is actually present in your situation, the way GOSH checked that a pit crew's single-leader, fixed-sequence structure applied to a surgical team before copying it.
The pit crew comparison did not occur to two surgeons because they were unusually creative. It occurred because they had already asked what their problem actually was, in words specific enough that a Formula 1 broadcast could answer it.