Trustees and providers will require robust approaches to AI, alongside clear governance frameworks with strong controls and human oversight, to stay in line with best practice, emerging guidance and standards when deploying AI, because the UK is taking a principles-based approach to AI regulation.
This means that individual industry regulators will provide guidelines for firms but otherwise trustees and providers will need to establish their own robust operating models, which include human oversight, and governance approaches.
This is significant given the centrality of AI reforms to implementing the swathe of pension reforms contained within the Pension Schemes Act, the newly launched Targeted Support regime and the further changes likely following the conclusion of the Pensions Commission.
Default retirement pathways, Value for Money assessments and small pots consolidation will all increase the need for providers to make better use of their data to support automated decision making and matching as well as standardised benchmarking.
This backdrop creates a clear expectation that providers will need to further their adoption of AI, but do so in a way that improves member outcomes while maintaining trust.
To achieve this, providers will need to create their operating and governance models within the boundaries of data protection rules and regulations, and to have the flexibility to adapt their models as regulatory guidance evolves, such as the guidance on the responsible adoption of AI expected later in 2026 from the Pensions Regulator.
Operating models will need to cover a range of AI techniques, such as clustering to identify patterns in data, classification to support consistent decisions, and ongoing monitoring to identify changes in behaviour and inform appropriate actions.
These models will need to ensure that all uses of AI follow a well governed path where human involvement is clearly defined - whether in the form of oversight activities, or through the incorporation of ‘humans in the loop’ to regularly handle and review outputs. They will also include policy controls, routing decisions and guardrails to provide a tightly controlled governance framework with defined intervention points.
Sami Saadaoui, Head of AI Architecture and Operations at Lumera, commented: “AI is set to become a critical enabler of the next phase of pension reform as the industry digests and begins to implement the Pension Schemes Act. Schemes and providers will need to leverage AI to deliver more personalised member outcomes, support automated processes at greater scale and improve the consistency of decision-making across increasingly complex datasets. However, the real challenge is not simply adopting AI, but deploying it within a robust governance and control framework. Pension providers and trustees will need clear accountability, strong human oversight and transparent decision-making processes to ensure AI is being used responsibly and in members’ best interests.
“The UK’s principles-based approach to AI regulation means firms cannot rely on prescriptive rulebooks alone. Instead, they will need to demonstrate that their operating models, controls and governance frameworks are sufficiently robust to manage risks around bias, data quality, explainability and consumer outcomes.
“The current swathe of reforms significantly increases the volume and complexity of data that needs to be processed and analysed. Firms need the scalable technology and human expertise to ensure that AI is unleashed to its full potential within defined guardrails. Those that manage this best will be best placed to capitalise on a new era of pension saving and access in the UK, delivering better outcomes and maintaining trust with members.”
|
