Board Toolkit
The four questions, and the checklist
Designed for use in a live meeting. Nothing here depends on technical knowledge.
VALUE
What measurable business problem are we solving?
Reject capability framing. Require a named problem, a baseline and a measurable outcome with an owner.
DATA
What data does this require, and do we have the right to use it?
Quality, ownership, lineage and lawful basis. Consent for one purpose is not consent for all purposes.
RISK
What happens when the AI is wrong?
Not if — when. Require the error profile, who bears the error, how it is detected and how it is reversed.
ACCOUNTABILITY
Who remains responsible?
A named executive. Vendors, models and committees do not absorb accountability. It remains human.
Board AI Checklist
Mark how well your organisation could answer each question today.
Where can AI create material value for us?
Reject 'AI opportunities' framed as technology rather than as business problems.
What data do we need, and do we have the right to use it?
Consent for one purpose is not consent for every purpose.
What should we build?
Building commodity capability is a common and expensive error.
What should we buy?
Check switching cost before checking price.
What should we partner for?
Partnership does not transfer accountability.
What risks are acceptable?
Absent an appetite statement, management sets appetite by default.
What must remain human-controlled?
This list should exist in writing before it is tested by an incident.
Who owns AI governance?
If no one can be named, no one is accountable.
How do we measure ROI?
Ask who verifies the benefit claim independently of the sponsor.
How will management report AI performance to the Board?
Reporting frequency should follow the risk level, not the calendar.