AI changes the questions directors need answered.

Understand the material use

Ask management where AI affects the organisation's business, people and risk. Identify consequential decisions, information practices and dependencies rather than receiving only a list of technologies.

The level of detail should support oversight. Directors do not need to become model engineers to ask whether a proposal has an appropriate evidence base.

Examine the basis of the decision

For material commitments, ask what alternatives were considered, what the benefit depends on and what could go wrong. Distinguish management assessment, vendor claims and independent evidence.

Where specialist knowledge is required, establish who provides it and the scope of their work. A broad assurance statement may not cover the actual deployment.

Clarify delegation

Management needs authority to run the business, with clear escalation for significant changes and exposure. The board should understand the roles and reporting arrangements rather than approve every routine use.

An AI committee does not, by itself, resolve accountability across existing executive responsibilities.

Test whether oversight reaches operation

Can management show controls in use, material incidents, unresolved findings and the response? Does reporting distinguish missing evidence from an absence of problems?

The objective is informed oversight, not documentation volume. A board minute is a record of discussion, not a guarantee about the underlying decision.

Keep legal and operational questions connected

Privacy, discrimination, consumer, workplace and sector obligations may be relevant in different ways. Their application requires assessment, not a generic claim that all AI uses carry identical duties.

ASIC's published governance work is useful context within its scope. Do not interpret a regulator's commentary as a new universal rule unless the legal basis supports it.