How much unverified AI work passed through your org last month?
Twelve questions. Get your Proof Gap Score, a dollar exposure estimate, and a breakdown across the four proof categories.
Payment Proof
Are AI-initiated or AI-recommended payments verified against duplicates and bank-detail changes before release?
Is every vendor bank-detail change independently confirmed out-of-band?
Can you produce, per payment, the evidence that controls were applied?
Decision Audit
Does each consequential AI decision capture a named human approver and timestamp?
Can you reconstruct the inputs an AI saw for a decision made months ago?
Are decision records tamper-evident (sealed / hash-chained)?
Output Verification
Are high-impact AI outputs checked against an objective standard before they’re trusted?
For AI-written code/work, do you verify tests actually ran and acceptance criteria were met?
Is the pass/fail result of each verification recorded?
Agent Reputation
Do you check an AI agent/vendor’s verified work history before relying on it?
Can you tell a verified agent from an unverified one at procurement time?
Do you require portable proof of past performance from AI vendors?