Founder note

Automate the routine. Preserve the practice.

When AI clears easy operational work, teams can get faster day to day and less ready for the rare outage. Write a monthly practice loop.

AI can remove a lot of routine operational work. That is useful.

Routine problems are also how people build judgment. If AI handles every easy incident, the team may get faster day to day while getting less practiced for the rare outage automation cannot save.

The founder move is not “use less AI.”

It is: automate the routine, and preserve deliberate human practice.

A simple monthly mechanism

1. List the incidents AI now resolves or assists.
2. Pick the failure modes humans still need to understand.
3. Once a month, run one realistic incident without leaning immediately on automation.
4. Rotate who diagnoses and who leads.
5. Record where human understanding thinned.
6. Update the runbook and the automation afterward.

Automation should remove toil without removing competence.

60-second drill card

System / service: ____________________
Month: ________  Practice hours this month: ____

What AI would usually clear: ____________________
Failure mode humans still need: ____________________
Who diagnoses / leads this round: ____________________

What broke our mental model: ____________________
One fix to docs / runbook / ownership: __________

Next practice date: ________

Honesty

This was sparked by living discussion about AI handling incidents and engineers losing touch with their systems, plus the classic automation irony: automation removes routine practice while leaving humans for the abnormal. I did not run a vendor’s incident simulators. Steal the practice rule, not a product demo.