[SPONSORED CONTENT]
AI is changing how members engage with their pensions, delivering answers at speed and scale and altering the relationship between provider and member.1 With more than one in three adults expecting to increase their use of AI for money management,2 these tools are likely to increasingly shape how people understand and make decisions about their pensions.
Who steers that conversation matters for member outcomes. People favour tools that are easy to use and always available; once habits form, they stick. That reality puts pressure on providers and scheme sponsors to act, to regain control of the conversation. One tempting option is to adopt a ready-made third-party AI engagement tool for members. But is this really the safest route?
The rise and risks of third-party AI
Third-party AI tools have become commonplace across industries. An MIT Sloan study of 1,240 organisations across 87 countries found that 78% of organisations use third-party AI. Yet, 20% admit they haven’t properly assessed the risks these tools bring.3
This issue is far from limited to other sectors. In financial services, AI use spans a wide range of areas, with operations and IT being the largest use case at 22%.4 About one-third of financial services AI use cases rely on third-party tools, and 46% of firms say they only partially understand how the AI systems they use actually work.5
This is potentially a significant concern, particularly given that 55% of AI failures are linked to poor use of third-party AI tools.6 Pension providers remain accountable for member outcomes regardless of where the technology comes from, so understanding and oversight are critical. In most cases, off-the-shelf AI is not designed around UK pension rules, scheme specifics, or regulatory safeguards. When systems that help shape member decisions are poorly understood, or weakly governed, member risks multiply.
Control, content and confidence
“We recognised that members are already using generalist AI to ask pension questions,” says Matt Gosden, Head of AI, Mercer Master Trust. “Ignoring this reality would leave them exposed. That’s why we developed a specialist pensions AI assistant that is now in-house, keeping governance and accountability close to the scheme.”
“Our AI pensions assistant draws on around 1,000 pieces of curated pension content, reviewed by experts and aligned with the Mercer Master Trust scheme rules. Testing and evaluation is continuous and rigorous. Pension experts actively challenge the AI, security specialists test its resilience, and extensive test libraries monitor quality and prevent regressions.
“By not providing regulated advice and limiting responses to curated, expert-approved content with clear oversight, we have addressed, and can mitigate, many of the risks associated with generic AI tools.”
What trusted, in-house AI looks like
- Clear boundaries and escalation paths so members can access human support when needed.
- Outputs that are validated against scheme rules, traceable to credible sources, and regularly updated.
- Privacy by design and strict data policies protect member information.
- Where building in-house AI isn’t feasible, adoption of third-party tools must be paired with strong governance frameworks, auditable decision trails and explicit ownership.
Accountability not just adoption
As AI becomes ever more embedded in the pension journey, the question is not whether to adopt AI, but how to do so in a way that protects and supports members now and in the future. Relying on third-party AI tools introduces governance and risk challenges that pension providers cannot ignore. Mercer’s in-house approach, set out in ‘AI, consumer expectations and pensions’, demonstrates that AI can be designed to support members. To work, it must be paired with strong governance, expertly curated content and clear scheme ownership to safeguard accuracy and not risk poor member outcomes.
To read more articles from Mercer, visit the content hub on Corporate Adviser – here.
____________________________________________________________________
1 Response to FCA Mills Review consultation – Quietroom.
2 Over 28 million adults now using AI tools to help manage their money – Lloyds Banking Group plc
3 Third-party AI tools are responsible for 55% of AI failures in business | ZDNET
4 Artificial intelligence in UK financial services – 2024 | Bank of England
5 Artificial intelligence in UK financial services – 2024 | Bank of England
6 55% of AI Failures in Business Attributed to Poor Use of Third-Party AI Tools