The Decision Changed. The Mandate Didn't.
Authority can move without anyone calling a meeting.
Most organisations know when they approve a new AI use case. They are less reliable at noticing when an approved use case quietly becomes a different decision system.
The movement is usually incremental. A recommendation becomes a ranking. A ranking becomes a default. A default becomes an automated action. A human still appears somewhere in the workflow, so the governance story sounds unchanged.
The most important authority changes may arrive as configuration changes, not committee papers.
This week's signals are not five new technologies. They are five ways the decision itself changes while the mandate appears to stand still.
The default path moved from review to automation.
Routine cases that were once presented to a human are now resolved automatically. Humans review only items the system marks as unusual.
The system no longer merely supports the decision. It decides which decisions humans are allowed to see.
Did the move to exception-only review trigger an explicit reconsideration of decision authority?
A new integration can be a new mandate in disguise.
An agent that could previously read and recommend can now write to a system, call an external service, send a message or trigger a workflow.
Execution capability expands. The gap between recommendation and consequence shrinks, even if the use-case label remains the same.
Which new tool classes automatically require a fresh authority decision rather than a routine technical change?
A right to challenge is not useful when there is no time to exercise it.
The human remains the formal approver, but operational speed has reduced meaningful review from minutes to seconds.
Approval survives on the process map while challenge capability weakens in practice.
At what review time does human approval become acknowledgement rather than judgement?
Same workflow. Larger consequence.
Value limits, confidence thresholds, customer groups, geographies or eligible actions expand inside an already approved workflow.
The organisation has delegated more authority without necessarily changing the name of the control or the accountable owner.
Which changes in scale, exposure or scope make the previous approval too small for the new decision?
The organisation can keep the same workflow while the decision engine changes.
A provider changes a model, capability, routing method or agent behaviour inside a service the organisation already approved.
The original authority decision may now rely on assumptions that are no longer true, even though no internal change request was raised.
What supplier-side change is material enough to reopen our decision rather than simply update the technical record?
The authority moved before the governance language did.
This is the problem with treating AI governance as a one-time approval exercise. The approved object moves. The decision path moves. The permissions move. Sometimes the accountability does not.
When the practical role of AI changes, the organisation should ask whether the authority decision has changed too. A material change in autonomy is not simply a technical event. It can be a reopening trigger.
Ask for the delta, not the policy.
| Ask | What a useful answer contains |
|---|---|
| What can the system do today that it could not do when approved? | Tools, data, actions, thresholds, users, geographies and execution rights. |
| What decisions do humans no longer see? | Cases now auto-resolved, suppressed, routed or handled by exception. |
| Where did review time shrink? | Processes where human authority exists formally but intervention is becoming impractical. |
| What changed outside our organisation? | Provider models, embedded capabilities, platform rules and supplier-side automation. |
| Which of those changes reopened governance? | A visible record of reassessment, not an assumption that the original approval still applies. |
Where has AI changed how the decision is made without changing who remains accountable for the consequence?
If the answer is “nowhere”, ask for one consequential process and compare the current workflow with the one originally approved. The difference is often more informative than the policy.
That gap is where authority drift hides.
Next: Board Case Study, 20 October 2026. The Decision Boundary.