ISSUE #2

The machine had a better idea

"Success breeds complacency. Complacency breeds failure. Only the paranoid survive."
Andy Grove

Last issue I confessed that an AI cloned our methodology and scored 68. This fortnight we turned the desk around. We asked a fresh AI for its own OKR doctrine, no input from us, and we marked its homework.

I expected to enjoy that more than I did.

THE IDEA

We graded it like a rival, position by position. On the big things we won, and not gently. Its core conventions, three objectives in parallel and scoring where 0.7 counts as success, are the exact machinery behind the OKR failure rate it politely warns you about. A doctrine that ships its own failure modes, confidently.

Comfortable so far. Here is the uncomfortable bit. Marking a rival where they are weak is not learning, it is reassurance. The red pen only earns its ink on the strong pages. So we graded it where it was strong, and on one component the machine was plainly crisper than us.

Grade the machine. Properly.

THE CHANGE

Its best idea: a clean lifecycle between KPIs and OKRs. A metric earns an OKR under exactly two triggers. It is broken beyond routine fixing, or you are deliberately pushing it to a new level. When the OKR lands, the metric returns to the scorecard with a new threshold. No permanent residents.

We half knew this. It sat in our methodology the way half known things do: implicit, unwritten, agreed with a nod. The machine said it cleanly, so we took it. Same day, into the methodology, credited in the changelog: adopted after benchmarking against the AI's default doctrine.

We also rejected its second best idea, in writing. Learning goals inside OKRs: thoughtfully designed, still wrong. The moment learning carries a grade, teams stop learning and start passing. Discovery stays outside OKRs. Taking what beats you only counts if you can also say no, with reasons.

THE QUESTION

Ask your AI for its doctrine on the thing your company does best. Then mark its homework the honest way: not where it is weak, where it is strong. Steal what beats you. Write down why you reject the rest.

The machine that copies you is also a free consultant, if you grade it instead of fearing it.

Grade the machine.

See you in a fortnight,
Matt

P.S. The free AI OKR Coach kit is built from the same methodology this issue just changed. Forward this to whoever runs your operating rhythm. They will know what to do with it.

← Previous issue All issues →
// that worked

That Worked

Short reports from the gap between plausible and works.

This story was an issue first. The next experiment is already running.

Fortnightly. Under 500 words. Every claim carries a receipt. Unsubscribe is one click and no hard feelings. Privacy. All issues.