LPPA-Burges Salmon survey: How pension members are using AI

Key takeaways from the survey

How members are using AI and planning to do so in the future

Whilst only 9% of members responded that they had used AI to look into pensions-related topics, around 58% of those people said that they either did (6%) change their pension as a result of the AI usage or would consider doing so in the future (52%).

However, AI is far from perfect and can make mistakes – especially where it doesn’t have all the information available (e.g. the provisions of a trust-based scheme’s rules, or information to confirm the LGPS regulations applicable to a particular member). Members may therefore be receiving misinformation regarding their benefits and rights, and then relying and acting on this. Schemes may wish to take action here to mitigate the risk to the member, and also to the scheme.

Members still want a human touch when it comes to AI adoption

Only 2% of members said they would be comfortable with AI making pension related decisions about them without human review, with 61% saying they would never be comfortable with this approach.

When the decision to be made about them involved human oversight before being implemented, nearly 23% said they would be comfortable now and just over 46% could be comfortable in the future – strongly suggesting that members would prefer for AI to support pension related decision making as long as a human remained responsible for the final outcome. This sentiment aligns with the PASA Guidance on AI, which states that AI should “augment, rather than replace” human decision making.

Some members are unaware of the cyber risk posed by using “public AI” tools such as ChatGPT

Namely, 39% of members responded that they did not know that “public AI” tools process and store data inputted into them, and train the tool based on that information too.

Given the cyber concerns regarding members inputting sensitive information belonging to them or to their pension scheme, it is striking that 39% of members are unaware of this risk.

Members would use AI to make complaints… and most would not declare doing so to the scheme

Around 50% of members said they either have (3.5%) or would (47%) use AI to make a pension scheme complaint or an Internal Dispute Resolution Procedure (IDRP).

Further, nearly 64% of members admitted that they would not declare the AI usage in the complaint document.

AI-produced complaints and IDRPs are a topic of interest in the industry at the moment, given the potential for the technology to significantly increase the volume of complaints/IDRPs received by pension schemes and funds, as well as increasing the length of the complaints and often containing substantive issues (such as inaccurate summaries of the law and not providing sufficient information to assess whether a “non-financial injustice” award should be paid).

Members would like help from pension providers about the risks of AI

Just over 75% of members said they would find it helpful for the administrator to explain the risks of using AI to make financial and retirement decisions, and nearly 81% saying they would find it helpful if the administrator were to provide “best practice” guidance on using AI in a safe way.

How we can help

We have produced a free “Pensions AI Tool” to help organisations take the first steps in developing effective AI governance. Access this here.

Please do reach out to Chris Brown (Partner), Callum Duckmanton (Solicitor), or your usual Burges Salmon contact if you would like to discuss this article further, including about how we can help.

Our comments