Why AI is harder in a business that already works
The constraints an established business arrives with — the process, the systems, the exceptions nobody wrote down — are not obstacles between you and the AI. They are the specification for it.
Insights
Ideas for making AI useful at work.
Notes on choosing the right AI work, building reliable agents, and helping teams use them.
Featured
Start with a real decision.
How to choose your first AI workflow
The usual selection method produces forty ideas scored on two axes nobody can estimate before building. Four questions predict a first workflow better, and every one of them is answerable in a week.
How to measure AI value without misleading the organization
Most AI metrics are true and useless: activity counted as outcome, projection reported as result, a pilot cohort extrapolated to a department. Five parts fix all three.
Human-in-the-loop is an operating model, not a checkbox
Saying a person reviews the output is not a control. A control specifies who reviews, against what standard, with what authority, and what happens when they disagree.
Why most AI pilots never become operating capabilities
A pilot proves a model can do something. Production proves an organization can rely on it. The gap between the two is not technical, and it is where most programmes stop.
Put the ideas into practice.
Have a workflow, product, or business problem you are considering applying AI to?