1. Go / No-Go Questions
Before Capital Commitment — Use these to validate whether an AI initiative is ready for material investment.
What is the one assumption that, if wrong, would make this investment worthless?
Exposes the single greatest capital risk—and whether it has been stress-tested.
What measurable business outcome will this AI initiative improve?
Boards care about capital outcomes, not internal metrics or technical milestones.
How will success be measured in business terms, not training metrics?
Aligns evaluation with real economic impact rather than statistical artifacts.
What is the baseline performance without AI, and how does this compare?
Without a baseline, improvement cannot be validated and investment cannot be justified.
What is the fallback or mitigation plan if the system fails?
Defines governance boundaries and loss containment before capital is at risk.
2. Continue / Stop Questions
When Progress is Ambiguous — Use these when technical progress exists but business value remains uncertain.
What have we learned that we didn't know at approval?
Focuses on evidence and insight, not activity or effort.
Has our confidence increased or decreased based on real evidence?
Separates genuine progress from optimism and narrative momentum.
What specific evidence would immediately justify stopping?
Forces clarity on exit criteria before sunk-cost thinking takes hold.
Are we measuring progress toward value or simply activities completed?
Distinguishes meaningful advancement from project motion.
Are there alternative uses of this capital that offer better risk-adjusted returns?
Introduces opportunity cost into the continuation decision.
3. Cross-Context Governance Questions
Relevant at Every Stage — Central to independent, fiduciary risk assessment.
Who has unambiguous authority to stop this initiative if outcomes deteriorate?
Clear delegation prevents escalation bias and diffused accountability.
What specific metrics will trigger intervention or termination?
Removes ambiguity in decision thresholds before pressure mounts.
How are learning inputs governed and approved at senior levels?
Learning is a capital decision—not a technical detail to be delegated.
How frequently will independent review occur, and what evidence will it require?
Governance cadence prevents unnoticed drift and internal bias.
What is the maximum acceptable capital loss before reevaluation?
Preserves capital by design, not hindsight.
4. Using These Questions
When to Use Each Section
- Section 1 — When approving new or scaled AI investments
- Section 2 — When projects stall or show mixed value signals
- Section 3 — As ongoing governance triggers in every review cycle
Boards that use these questions systematically avoid costly escalation driven by proxy metrics, escalation bias, technical optimism, and unstructured progress reports.
These questions support capital stewardship—systematically, independently, and transparently.