domain
Corporate Governance
Board-grade oversight of what the AI fleet actually does
Boards are now asked to oversee AI usage the way they oversee financial controls, and most organisations can offer them only assurances. CueCrux offers them attribution: what the fleet did, agent by agent, verb by verb, with a signed receipt behind each count and nothing estimated. "Your agents did 4,183 things today" is a reportable fact, not a vibe.
Delegated authority gets enforced rather than described. Risk classes on plans and work items determine what agents may do autonomously; anything above the line routes to a human gate that cannot auto-approve past a timeout, and each approval is a signed fact naming the approver. That is a directors' duty of oversight translated into machinery.
Decision records preserve the reasoning behind non-trivial choices, so governance reviews and post-incident inquiries read a contemporaneous log rather than commissioning an archaeology project. Cost attribution closes the loop for audit and remuneration committees: the spend on automated work reconciles to receipts, including every metered credit.
For company secretaries and general counsel, the practical change is that board reporting on AI stops being a survey of what teams say they do and becomes a query over what provably happened.