A quiet revolution is happening in the pensions industry, as it starts to embrace the use of artificial intelligence (AI) for more than number crunching.
AI has been used in pensions for several decades, but this has largely been for back office work, such as analysing large data sets associated with pension schemes, which have thousands of members and billions of pounds of assets.
Opportunities and concerns
The advent of generative AI brings greater opportunities. Pension scheme members are using AI in their daily lives and the industry needs to keep pace. Both major industry regulators - The Financial Conduct Authority (FCA) and The Pensions Regulator (TPR)
- are supportive of pension scheme initiatives involving the use of AI. The regulators also employ AI in their own work.
TPR recently reported that “a coalition of fraud fighters”
had used AI to help build a new tool to detect pension scam websites, which would have had the potential to reach thousands of pensions members, putting their retirement savings at risk.
However, the pensions industry needs to overcome some concerns before AI is more widely adopted in its interfaces with scheme members. The Treasury Select Committee recently issued a call for evidence on “The
use of AI in banking, pensions and other financial services”. To inform its response to this call for evidence, the Society of Pension Professionals (SPP) used their 2025 AI
survey of its members, which had assessed the current use of AI in the pensions industry, how this is expected to change in the future and to identify any barriers to more widespread adoption.
87% of pensions professionals who responded to the SPP’s 2025 AI Survey said that their firm uses AI, but 75% said that this is currently limited to only 1-5% of their services. It was widely acknowledged that AI saves time and reduces cost, but organisational
nervousness was identified as the biggest barrier to firms adopting more widespread rollout.
According to the SPP survey, within the next 10 years, the use of AI in the pensions sector is expected to increase significantly, with 18% of respondents expecting it to be used in more than 50% of their services. This implies some optimism that operational
nervousness will lessen with the passage of time.
The member engagement conundrum
The use of AI to aid member understanding and engagement, particularly in the years leading up to retirement, is a hot topic at present. Some background information might be useful to understand why member engagement is a problem.
More than 11 million UK workers have been automatically enrolled into a pension scheme since new employer duties were introduced in 2012. Automatic enrolment relies on member inertia (i.e. members have to take action if they wish to opt out of a pension
scheme, and if they do, they are automatically re-enrolled three years later). Essentially, it is easier for the UK workforce to be in a pension scheme, rather than not to be in a pension scheme.
Most of these automatically enrolled members are in Defined Contribution (DC) schemes, where the employer contributes a fixed percentage of payroll but the members bear the risks of assessing whether their contribution levels and investment choices will
support the level of income they need in retirement. At the point of retirement, there are difficult choices to make.
For example, members need to make decisions around if and when to purchase an annuity, whether to take drawdown income, or whether to take the whole pension pot as a lump sum (with 25% being tax free). These decumulation choices are complicated, but most
members do not take financial advice. The FCA has issued proposals to provide more targeted support for scheme members, citing a survey in which “75% [of scheme members] did not have
a clear plan for how to take their money out of their pension or didn’t know they had to make a choice.”
Further provisions will be included in this year’s Pension Schemes Bill requiring trustees to provide decumulation pathways, but it is likely to be a few years before these new duties are fully in force.
How is AI helping?
There is a lot of potential for using AI in helping to guide DC members through various stages of their pensions journey. These include personalised communications or video clips to prompt members to consider if they are saving enough – this could be powerful
if targeted at key birthdays, for example, or to coincide with an annual pay review. Greater technology is also being deployed to guide members through decumulation processes.
The pensions industry is making technological strides in member support - particularly Master Trust Schemes and Pension Providers. This is acknowledged in a recent
report by the Pensions Policy Institute (sponsored by TPR), which acknowledges a number of “personalised, engaging, and practical” platforms that have developed to help individuals to understand and plan for their retirement. These include interactive tools
and video content. But, as the report concludes, “Considerations around regulation and consumer protection will be crucial in ensuring these tools serve savers’ best interests, striking the right balance between accessibility, reliability and safeguarding
against risks.”
Looking to the future
Regulatory clarity, particularly the work currently being undertaken by the FCA around targeted support, will hopefully provide greater certainty for the industry and reduce nervousness that new technology may stray unwittingly into regulated areas. This
in turn should lead to greater confidence in investing further in AI based pensions technology.
Pension scheme trustees need to additionally consider how any technological developments sit with their fiduciary duties. They need to make sure that advancements are in the best interests of scheme members, that the pace of change is appropriate for different
demographics and that any AI used by third party service providers is subject to good oversight to minimise errors. Trustees should ensure that they have robust, GDPR compliant agreements in place with service providers using AI, and need to consider, for
example, whether data protection impact assessments are required and whether privacy notices need to be updated.
However, it is unclear if the general public are ready to trust AI to assist with important financial decisions. Generation X joined the workforce when desktop computers were a novelty and when the main form of external communication was a telephone conversation.
Are they ready to now allow AI to help them with their retirement choices?