The Blackout Test as Governance

I didn't need to wait for the result to have my diagnosis.

The resistance was immediate. Not technical resistance, not infrastructure or operational risk resistance. It was the resistance of people who no longer know how the process works without AI in the middle.

That has a name: implementation without internalization.

The company integrated the agent, the process became dependent on it, but nobody learned what the agent does, why it does it, or how to review what it delivers. The black box became a crutch, and the crutch became a wall.

The blackout test is simple to propose: take a critical process that currently runs with AI and ask the team to operate without it for 24 hours. Not as punishment. As diagnosis.

If the team can manage — even with more effort or more slowly — the AI is well positioned: it accelerates something the team understands and controls.

If the team freezes, panics, or rejects the test before even trying, you don't have an operational continuity problem. You have a governance problem that no system audit will reveal.

Resistance to the blackout says more about institutional absorption than any usage report or automation dashboard.

The question was never whether the company uses AI. It's whether the company knows what it's using — and whether it survives when it stops.

Could you run your critical processes for 24 hours without AI? Comment here. I want to understand where this gap hurts most in your operation.

And save this post: the blackout test may be the most honest governance tool you'll apply this year.