AI Literacy Knowledge Hub

What Evidence Should Boards Request When Staff AI Training Is Marked Complete?

Quick answer

Boards should ask for evidence that staff can apply AI guidance in their roles, alongside completion figures. A useful report shows who was trained, which tasks were assessed, how errors and escalations are monitored, and what changed after weak results. The board should also ask where evidence is thin and who owns the next action.

What to remember

Key takeaways

Completion measures reach, not safe application.
Request role-based evidence and exceptions, not just a single percentage.
Ask how incidents or complaints influence the learning plan.
Track owners, actions and repeat evidence over time.

A board pack may report that almost everyone completed AI training. That is useful for coverage, but it does not show whether staff can recognise an unsuitable use, check an output or escalate a concern. The board’s role is to challenge whether the evidence supports the assurance being offered.

Ask what the number represents

Separate enrolment, completion, knowledge checks and observed performance. Request the population covered and excluded, including relevant contractors where the firm’s programme includes them. Understand which tools and roles the assessment actually tested.

Use existing board assurance habits

Boards already challenge control attestations, customer outcomes and operational incidents. Ask for a small set of role-specific examples, assessment results, quality reviews and significant exceptions. Invite the accountable owner to explain a weak result and the corrective action.

Where AI helps

Approved analytics may group assessment patterns or flag repeated questions for review. A dashboard can make gaps easier to see, but it cannot decide whether a training exercise reflects real work. Human owners must interpret the evidence and validate data quality.

Make the question actionable

Ask who owns the programme, who independently tests the evidence, what has changed since the last report and when the board will see a follow-up. NIST’s AI RMF links role-specific training, documented responsibilities and monitoring; it is voluntary guidance, not a UK board reporting rule.

Example

A London retail lender reports full completion of an AI awareness module. The board asks for evidence from customer service and credit review roles. A sample exercise finds that some staff cannot identify an unsupported AI summary. The sponsor sets targeted practice and reports the next quality check to the board.

FAQs

Is a completion rate useless?

No. It shows reach, but needs to sit alongside evidence that people can apply the learning.
Should the board inspect individual staff scores?

Usually a role-level picture and material exceptions are more useful; follow the firm’s governance and privacy arrangements.
What if the results are poor?

Ask for the affected tasks, immediate controls, accountable owner and repeat assessment date.