The approval was right when it was made.
The most dangerous decision may not be the wrong one. It may be yesterday’s right decision, carried forward after the facts have changed. This article looks at stale approvals, why they become dangerous in AI-driven workflows, and why good governance must know when to stop, reassess and ask again before consequence.
Bob McTaggart edited with ai
9/5/20261 min read
The approval was right when it was made.
By the time it reached us, it wasn't.
I learned that lesson long before AI.
During my service, I saw a decision arrive late and based on circumstances that had already changed. The decision itself was not foolish. Based on the information available when it was made, it made sense.
The problem was time.
Events had moved on.
New information existed. Conditions were different. But the approval was still travelling through the system as though nothing had changed. Had nobody stopped and questioned it, the consequences could have been disastrous. That stayed with me.
An approval is not necessarily a permanent permission. It belongs to a particular set of facts, at a particular moment, for a particular purpose.
Change those facts enough and somebody should have to look again.
We understood that with people long before AI came along.
Now imagine an AI agent working through a process for hours, days or weeks.
A human approves something near the beginning. The data changes. The customer changes something. New information arrives. The risk changes. Should that original approval still be enough?
I don't think so.
That question became part of the thinking behind our patent-pending work: when does an earlier approval become stale, and what change should force the system to stop and ask again?
AI didn't create this problem.
It inherited a very old human one.
Sometimes the dangerous decision isn't the wrong decision.
It is yesterday's right decision being acted upon today.
That is the governance point.
Human approval should not be treated as a permanent green light. Good governance has to know when the facts have changed enough that the old approval no longer applies — and the system should stop before consequence, not explain itself afterward.
Supporting
Getting Veterans and First Responders back on mission.!
Veteran-inspired AI Governance & Trust Infrastructure
Trusted by Heroes and Mounted Rifles Management
Leadership and peer support are taught through RedFridayTalks.Help
The same governance protections are available to everyone.
© 2026. All rights reserved.