Your Domain Is Your Edge

A lot of my mentoring conversations have turned into the same one. What should I be learning to keep up with AI? Am I already too late? What can it actually do for me? Underneath most of them is a quieter worry: is the thing I’m good at about to stop mattering?

My answer, for nearly everyone, is the opposite of what the worry assumes.

AI makes the medium cheap

Think about what’s actually getting cheaper. Writing code, drafting documents, summarising, translating between formats: the medium through which expertise gets expressed. What isn’t getting cheaper is knowing what’s worth expressing.

Take the quant engineers I work alongside. They write a lot of code, but code is just the medium for their mathematics. What they really bring is the maths itself: pricing models, a feel for when a number is wrong, the ability to read a new research paper that most engineers couldn’t follow and that an AI would struggle to apply correctly. That’s scarce. The code that implements it is increasingly not.

So the advice isn’t “learn to code faster than the machine”. It’s double down on your domain, because that’s the part that’s becoming more valuable, not less. The same applies to anyone whose expertise sits behind a tool: a support specialist who knows exactly why customers get stuck is better placed to design a helpful assistant than an engineer who’s never taken a support call.

It isn’t too late

Everyone is early. The tools changed substantially this year and will change again next year; nobody has ten years of experience with them. What matters is the habit of learning alongside them, which is the same thing I wrote about in Your Job Is Not Your Identity: your role is what you’re focused on now, not a fixed identity, and skills carry over.

Use AI as a lever on your edge

The useful question isn’t “will AI replace what I do?” but “what does AI let me do more of?”:

• Prototype faster: try three versions of a model or an idea in the time one used to take.
• Offload the plumbing: data wrangling, boilerplate, test scaffolding, format conversion.
• Understand unfamiliar code quickly, so you’re less blocked on other people’s systems.
• Draft first, judge second: let it produce the first version, and spend your time on whether it’s right.

The pattern is the same one I like in platform design: let the domain expert define the behaviour, and let the tooling do the rendering (Keep the UI Dumb: Let Domain Experts Publish Schemas). AI is becoming that tooling for everyone.

The skills that make the transition smoother

Four skills make AI leverage safe and productive, whatever your domain:

1. Specifying clearly. Examples, constraints, what’s out of scope. The quality of what you get back is mostly the quality of what you asked for.
2. Verifying critically. Tests, sanity checks, reading output as a sceptical reviewer. Fluent output is not the same as correct output, and your domain knowledge is exactly what lets you tell the difference.
3. Thinking in systems. How the piece fits with everything else: contracts, dependencies, what breaks if it’s wrong.
4. Explaining to non-experts. As more work is done with tools, the people who can translate their expertise for others become the hubs.

Where to start

Pick one recurring task a week, something you already know well enough to judge, and try doing it with an AI assistant. Keep a short note of what worked and what didn’t. Within a month you’ll have a much better sense of where it helps you than any course could give you, and you’ll have done it from the safety of your own expertise, which is exactly where the edge is (Working with AI Coding Agents has more on working well with agents).

The world is moving towards valuing what you know over how fast you can type it. That’s good news for anyone who has spent years getting deep in a domain.