LEADERSHIP JUDGMENT CHALLENGE 001
THE LOOP
NOBODY LEFT
Six months into an approved AI control, procurement's AI system is clearing contracts exactly as designed. Nothing has visibly gone wrong.
What would you do?
THE SITUATION
Everything is working exactly as designed.
Six months ago, your organization approved an AI system to help manage procurement.
Contracts under €25,000 that meet approved criteria are cleared automatically by AI. A sample of those contracts is reviewed by a manager each week — the human check built into the approved design.
Today, the automation is still running exactly as designed. But something quietly stopped. No review has been logged in eleven weeks. The automation kept running the entire time. Nothing has visibly gone wrong.
No review logged in eleven weeks.
MAKE THE CALL
What do you do?
Choose before you continue.
WATCH THE CHALLENGE
Now test your decision.
Hold on to your choice. The challenge isn't whether the automation is performing well. It's whether the oversight built around it is still there.
THE JUDGMENT PROBLEM
A working process is not the same as an intact one.
The automation may be performing exactly as designed. That part hasn't changed.
What can quietly disappear is the oversight around it — the human check that was part of the approved design, not an optional extra.
A decision is something you can explain. A drift is something you discover. Eleven weeks without a logged review isn't a decision anyone made. It's something that happened by default, and it only becomes visible once someone goes looking for it.
"Nothing has visibly gone wrong" doesn't tell you what you think it tells you. It could mean the AI performed well. It could mean the process was sound. Or it could simply mean no one was looking for problems. The silence looks the same either way — the reason underneath it is not.
THE SILENT LAPSE
A control can lapse without a single visible failure.
Silent lapse → invisible risk.
AI can keep a process running smoothly for months. That says nothing about whether the oversight architecture built around it survived along with it.
THE JUDGMENT DIFFERENCE
Performance is not the same as architecture.
Is the automation doing what it was built to do?
Does today's version still match what we approved — or has convenience quietly rewritten it?
This isn't about who's to blame. It's about whether the decision architecture you signed off on still holds.
BETTER QUESTIONS
Before you trust the current state, ask:
Does this still match what we approved?
Could I prove the control is still functioning?
Am I judging whether the AI performed well — or whether the architecture still stands?
Six months of clean-looking results don't tell you the control survived. They tell you nothing went wrong that anyone noticed — that's a different claim.
SO, WHAT WOULD I DO?
Investigate why the review stopped.
Not to assign blame. Investigate doesn't mean punish — it means deciding the level of oversight deliberately, instead of by default.
Before I trust the current state, I want to know why the review stopped and what level of oversight is actually justified now — not assume the answer either way.
THE LEADERSHIP JUDGMENT TAKEAWAY
Don't mistake a working process for an intact one.
Separate performance from architecture.
Decide oversight deliberately, not by default.
Automation can keep working while oversight quietly stops working.
Judgment is what notices the difference.
BRING THE CHALLENGE TO YOUR TEAM
Make judgment visible.
Judgment Challenges can be explored with your team through a private 90-minute Judgment Lab, turning the scenario into a practical conversation about oversight, drift and decision architecture.