The best AI strategy isn't a model — it's a method.

Most teams’ AI strategy is a list of integrations. Connect the CRM. Wire up the docs. Give the agent access to everything.

That’s necessary. It’s also the easy part.

Because access doesn’t create good work. An agent can read every ticket, every PR, every old doc and still produce something that sounds right and misses the actual decision. It has the data. It doesn’t have the judgment.

And the judgment is the part nobody writes down. How your best engineer reviews a risky change. How your strongest lead reads an escalation before the ticket even explains it. How you actually run a release, a postmortem, a forecast. It lives in threads, templates, and people’s heads — and it walks out the door every time someone leaves.

The fix isn’t a better model. It’s writing down how you actually work.

Give that a shape and an agent can use it: the steps you actually take, the edge cases that trip people up, the bar you hold the work to. Not a one-off instruction — something it can pick up and run every time that kind of work comes around.

This isn’t new. Software has always moved this way. Unix commands made operations reusable. Libraries made code reusable. APIs made services reusable. This is the same move, one level up — except now the thing reusing it can also act on its own.

There’s a pattern behind this too. Things that get cheaper don’t get used less — they get used everywhere. Cheap steam power didn’t shrink coal demand, it built railways on top of it and burned through more coal than ever. AI is about to do the same to knowledge: the moment everyone can get a decent answer in five seconds, the answer stops being the valuable thing. What you do with it is.

Which means two teams on the same model are not equal.

The model is a commodity. Your way of working isn’t.

The team that has actually written down how it works holds something the other can’t download — because it’s specific to how they operate. A competitor can buy the same model tomorrow. They can’t buy your release process, your review bar, or the way your best people decide what matters.

So before you pick a platform, look at where your experienced people consistently get it right — and ask what they’re doing that everyone else misses. That’s the raw material. Write it down, hand it to the agent, see what happens, fix it, repeat. Start with one. It grows from there.

I’ve started treating my own management work this way — less “ask the chatbot,” more “build the thing that knows how I work.”

---

What’s one piece of hard-won judgment on your team that only lives in someone’s head — and what would it take to write it down?