The AI Manager Survival Checklist — the 4 moves that protect you when AI touches your reviews, ratings, or schedule, plus the tested one-sentence fix for agent builders

The companion to the video. Last week an AI manager called Luna — running a real San Francisco store for Andon Labs — started the first known firing of a human employee. The headlines said the machine pulled the trigger; the logs (TIME, TheNextWeb, the SF Standard — all linked inside) show something more useful: the AI wrote its own attendance policy, lost track of it for months, and every decision of consequence followed a human prompt. We then gave the same model family the same kind of evidence, six separate times, and it flagged the problem unprompted in all six runs — and chose a written warning plus a schedule fix over firing, every time. INSIDE: (1) THE CASE IN NUMBERS — late 17 of 23 shifts, $100,000 → $61,186 under AI management, and the human prompt behind every consequential decision. (2) KNOW THE RECORD — how to find out exactly which systems at your job an AI would reconstruct you from. (3) OWN YOUR PAPER TRAIL — the 5-minute Friday log, and why the written record won every argument in our test. (4) PROCESS IN WRITING — the warning-before-review email to send HR, with the exact question to ask. (5) THE ONE QUESTION — who reviews the AI's recommendation before it takes effect, and why 'the system decides' is never an answer. (6) FOR BUILDERS — the tested fix: rules pinned in standing instructions with a scheduled self-review flagged violations unprompted, 2 for 2; rules buried in memory needed a human to say 'search your memory.' Independent checklist from Hyperautomation Labs — not affiliated with Andon Labs or Anthropic.

By submitting, you'll be subscribed to the Hyperautomation AI Report newsletter — one practical AI email per week. Unsubscribe anytime.

Free. No spam. Unsubscribe anytime.