AI in pensions: From emerging technology to trustee responsibility 

Providers are using AI in administration, member communications, cyber security, investment processes and internal operations. For trustees, the question is no longer simply whether AI is being used. It is whether its use is understood, proportionate and properly governed.

Trustees do not need to become AI specialists, but they do need a structured way to ask the right questions. A practical AI workstream can provide that structure.

It might include a Trustee checklist covering use cases, data, model risk, accountability and oversight; an AI policy setting expectations and guardrails; and a questionnaire for providers to explain where AI is used, what risks arise and what controls are in place. It could also consider carefully controlled applications that improve efficiency, such as using AI to support the drafting of meeting minutes, while retaining human review and responsibility for the final record.

Oversight through the existing governance structure is important. AI should not sit in isolation as an “innovation” project. It connects directly to existing trustee responsibilities: data governance, cyber risk, outsourcing, business continuity, member communications and operational resilience.

A useful principle from recent pensions thinking on AI governance is that the level of control should be proportionate to the risk. A tool used to summarise a public document is not the same as an automated process that affects a member’s benefits, retirement outcome or access to payments. The latter requires much stronger controls, meaningful human oversight and clear accountability. In some sensitive decisions, automation may simply be inappropriate.

It is also worth creating an AI register: a living record of use cases, owners, risk ratings, approvals, controls and review dates. This creates visibility and helps trustees avoid discovering AI use only after a problem occurs.

AI can improve how pension schemes operate and, ultimately, how members experience them. But trust will depend less on how quickly schemes adopt the latest tool than on whether they can demonstrate that it is being used responsibly. For trustees, good AI governance is not about slowing innovation. It is about creating the confidence to use it safely.