The situation in the room
A board receives a polished AI investment proposal with competitive context, an efficiency claim, and a clear ask. The discussion stays general and the budget passes without a named owner for failure or adoption.
The proposal is strongest where boards are used to looking and thinnest where AI programmes fail: ownership, data, people, alternatives, evidence, recourse, and the path after a pilot. The work looked active from the outside. Inside the organization, the unanswered decision was still controlling what happened next.
1. What problem are we solving?
Describe it without naming the technology. If the problem disappears when the word AI is removed, the proposal may be a capability looking for a justification.
Ask what the organization would do without AI and why that alternative was rejected.
2. Who owns the outcome?
Approving the budget also approves the organization's current answer to failure.
Ask what authority that person has to stop, change, or narrow the system.
3. Who is affected?
Identify customers, employees, applicants, partners, and people indirectly affected. Ask who was consulted before the proposal reached the board.
A general stakeholder category is not evidence of consultation.
4. What data does it touch?
Ask what personal, confidential, or regulated information enters the system, where it is processed, how long it is retained, and whether it improves a vendor model.
Require a specific privacy and security assessment rather than a statement that the vendor is compliant.
5. Can we explain and correct a decision?
Ask how a person receives an explanation, who provides human review, what evidence is retained, and what recourse exists.
If the organization cannot reconstruct the decision, the board is accepting that limitation now.
6. What happens after the pilot?
Name the decision-maker, decision date, scale conditions, production budget, and stop criteria before the pilot begins.
A demonstration with no path to operations is not a transformation plan.
7. What does adoption require?
Look for budget and protected time for workflow design, role-based learning, support, and manager follow-through, not only licences.
Attendance at training is not evidence of changed capability.
8. How dependent are we on the vendor?
Ask what happens if terms, pricing, ownership, model behaviour, or service availability change. Understand export, portability, and exit.
A young market makes switching conditions part of governance, not procurement detail.
9. How will we know it worked?
Agree on the baseline, outcome, evidence, timing, and counterfactual now. Separate released capacity from cash savings and quality from speed.
Do not let the success measure change after the results arrive.
10. What did we learn last time?
Ask which prior technology or AI initiative is most comparable and how this proposal reflects its adoption, governance, and scaling lessons.
Organizations that cannot remember their pilots are likely to repeat them.
What good practice looks like
Tie approval to a named outcome and a fixed follow-up date. Review against the measure agreed before spending, not a new story constructed afterward.
Good practice is visible in ordinary work. People know what they may do, what they must check, who can decide, and where an exception goes. Leaders hear the same account from different teams because the framework is shared rather than remembered differently.
The aim is not perfect control. The aim is a system that notices when reality changes and knows who has to respond. That creates movement without pretending uncertainty has disappeared.
Make the answer operational
Start with the smallest version of the decision that still changes behaviour. Put it in plain language and test it with the people who will use it under pressure. Ask them what they would do, what they would record, and who they would call when the ordinary case stops being ordinary. Their hesitation will show where the instruction is still abstract.
Then connect the decision to the systems around it. A rule without training becomes a document people forget. Training without authority becomes advice people cannot follow. Authority without review becomes a permanent answer to a changing question. The three have to move together, with one person accountable for keeping them connected.
Review the result through actual work rather than confidence surveys alone. Look at a sample of decisions, outputs, exceptions, and escalations. Ask whether the intended boundary held and whether people found a hidden route around it. When the evidence changes, revise the operating choice openly and record why. That is not inconsistency. It is governance doing its job.
What to do on Monday
1. Put ten governance questions into the next board paper template. Keep the action small enough to complete, specific enough to observe, and connected to a named decision rather than a general ambition.
2. Require a named owner for outcomes and failures. Keep the action small enough to complete, specific enough to observe, and connected to a named decision rather than a general ambition.
3. Schedule the evidence review at the time of approval. Keep the action small enough to complete, specific enough to observe, and connected to a named decision rather than a general ambition.