Automation & AI

Demo or habit: what makes it stick

Most experiments die not because the tool failed, but because it never became something anyone would miss.

The pattern repeats everywhere: a brilliant demonstration on Monday, genuine enthusiasm, and by Friday nobody has opened it again. The tool worked. It just never became part of how anything actually gets done.

Attach it to a routine that exists

The single most reliable predictor of whether an automation survives is whether it sits inside an existing routine or beside it. If someone already replies to enquiries, the drafted reply has to appear inside that step — not in another tab, another tool, another habit they have to remember.

You are not asking anyone to use AI. You are making the tool the fastest way to do what they already do. If it is not the fastest way, it will lose, and it should.

One owner, one number

Two rules do most of the work. Every routine gets a named owner, not “the team” — shared ownership means nobody notices when it stops. And every routine gets one metric: hours back, first-reply time, quotes drafted. One. If you cannot name the number, you cannot tell whether it is working, and you will end up arguing about vibes.

Expect to retire most of it

Treat what you build as a portfolio. Most of what gets made in an enthusiastic month will not be used in the third month, and that is not failure — it is the cost of finding the two that matter. The discipline is being willing to switch off the ones that did not stick, rather than keeping them alive out of sunk cost and calling it a stack.