
What A Real Decision Record Looks Like
Most human-in-the-loop AI systems produce logs, not decision records. A log tells you a button was clicked. A real record tells you what the human knew, decided and what happened.
Writing
Long-form thinking on AI product development, leadership, and the craft of building things that matter.

Most human-in-the-loop AI systems produce logs, not decision records. A log tells you a button was clicked. A real record tells you what the human knew, decided and what happened.

AI agents can make technically valid recommendations that are still wrong because they lack business context. More data does not solve it. Better context does.

Human-in-the-loop AI does not automatically create meaningful oversight. If the interface quietly turns review into queue-clearing, what you have is approval theatre.

The serious money in AI isn't hiding in the shiniest product. It's sitting in the invoice queue nobody owns, the compliance check that still depends on someone reading three documents, and the procurement exception that delays a supplier payment.

For years, UX fought to get invited earlier. AI may have finally kicked the door open — but it also changed what we are expected to bring into the room.

The question is not 'should we train our own model?' The better question is: what will our product learn that nobody else can see?

The best AI strategy session may start less like a hackathon and more like a good piece of UX research: who is doing the work, what are they trying to decide, and where does the system currently fail them?

I threw myself into building AI apps and came out with an uncomfortable conclusion: the app was often the least interesting part.

'Where should we use AI?' is too broad a question. It usually produces a brainstorm, a backlog, and a few ideas that sound impressive until someone asks who owns the data.

The easiest AI product to build is often the easiest one to copy. Here's what separates a fragile wrapper from a durable workflow AI product.