Adviser — Data and AI Governance

Start with actual uses

Map relevant data flows and use cases, including informal adoption and supplier-provided features. Identify who owns the purpose, data quality, testing and human review of each use. Distinguish low-consequence assistance from applications that require stronger assurance or specialist legal assessment.

Advisory responsibilities

Recommend approval and review processes proportionate to the use case. Define expectations for documentation, evaluation, monitoring, user communication and escalation when outputs are unreliable. Examine supplier terms and the organisation’s ability to understand, challenge or discontinue a service. Include procedures for changes after initial approval.

For each material use case, recommend a clear owner for deciding whether outputs are acceptable and a route for users to challenge or correct them. Evaluation should reflect the actual population, language and service context rather than a supplier demonstration alone. The adviser should ask what happens when the system changes, the input data deteriorates or staff begin relying on it for a purpose beyond the approved use. Governance must address those changes over time, including the ability to suspend use when evidence is inadequate.

Qualified candidates

Candidates should combine data-governance experience with practical understanding of AI systems and their limitations. Work in education or public-interest services is valuable. They need to translate technical and policy issues into usable operating responsibilities, with specialist advice sought for applicable legal requirements.

A useful outcome

The engagement should leave a prioritised use-case register, clear accountability and a workable review process. It should not promise that a policy document makes every application safe or compliant. Candidates should discuss an AI use they advised against or constrained because the evidence was inadequate. Relevant supplier and advisory conflicts require disclosure.