Writing

Field notes on leading the shift.

What I learned going from consultant to operator to builder, and what it means for how you lead in the AI era.

Your sent folder is already a training set

You don't have to describe your tone to AI. You've already written thousands of examples of it.

The dangerous output is the confident one. The useful one tells you where to apply your own judgment.

The first pitch for AI is always headcount saved. The bigger story is the work you couldn't do before at all.

AI can produce the work. It can't supply the judgment about your business that makes the work right. That part is yours.

Scattered AI use feels like adoption. Often it's five people solving the same problem alone, five different ways.

Most prompts come back thin because they're missing a piece. Name four, and the work comes back finished.

High adoption feels like progress. The number that counts is output tied to a business objective, at finance-grade accuracy.

The risk in finance isn't a bad prediction. It's ungoverned data moving through systems nobody approved.

Execution looked like the edge. It never was. The skill now is knowing what good looks like, and directing the work to it.

The bottleneck with AI is almost never the model. It's whether you have access to data worth thinking about.

Most leaders ask whether AI can do the job. The question that moves you forward is whether you should, and how you will direct it.

I advised companies on change, then had to execute it, then build it myself. Each seat taught me something the one before couldn't.

The hardest calls in AI aren't technical. They're business judgment, and they're yours to make. You can't make them without building.