Pinsent Masons partners Charlotte Scholes and Hayley Goldstone explain how members are using AI to assist with pension complaints, and how trustees can respond appropriately – potentially using AI themselves.

There is something quietly significant happening in pension complaints.
Members who once submitted handwritten letters of complaint to trustees or pension managers now submit multi-page complaints that reference regulatory frameworks, cite apparently authoritative case law, and deploy legal language with a fluency that bears little resemblance to their actual familiarity with pensions law.
The common thread is artificial intelligence (AI), and its arrival in the complaints ecosystem raises questions that the pensions industry has barely begun to answer.
Without doubt, AI can improve access to justice. A member can now describe their situation conversationally, ask a large language model to help structure their complaint, and receive a coherent, organised submission within minutes at no cost. For members who are elderly, unfamiliar with formal correspondence, or simply cannot afford legal advice, this levels the playing field tremendously.
However, in the authors’ experience, the same capability that evens things out for members also serves to inflate disputes, making them bigger and more complex. It also risks flooding schemes and the Pensions Ombudsman (and potentially the Financial Ombudsman) with a tidal wave of complaints.
How AI has created an imbalance

AI systems, which are optimised to be comprehensive, tend to itemise every conceivably relevant legal argument, recite principles that do not apply to the facts, and repeat themselves across submissions to different bodies.
A member with a straightforward complaint about a delayed transfer value quotation may submit a document alleging breach of fiduciary duty and violations of the disclosure requirements under the Occupational and Personal Pension Schemes (Disclosure of Information) Regulations 2013. Meanwhile, the core grievance – a delayed transfer – is in danger of being buried in a wall of legal verbiage.
Unfortunately, trustees and their advisers cannot simply ignore all of these aspects of the complaint. They need to respond (albeit proportionately) to the member or risk an adverse finding by the Pensions Ombudsman.
The increased time and professional fees involved in responding to member complaints create a structural imbalance in the complaints system. Generating a sophisticated 2,000-word complaint takes a member perhaps 20 minutes. Investigating it properly, responding to each substantive point, and briefing legal advisers may take schemes days.
The statutory Internal Dispute Resolution procedure was designed for a world where the burden on each side was broadly proportionate and equal. That assumption no longer holds.
The effects of AI-generated complaints

While an AI-generated complaint may appear sophisticated, it will not always have merit. Schemes and their advisers should adopt a disciplined approach, strip away the legal language, and ask: what is this member actually saying happened, and are they right?
In addition, dismissing a complaint because its AI-generated legal citations are wrong is not sound practice. It would sit uncomfortably before the Pensions Ombudsman, who will consider complaints of maladministration irrespective of how articulately they are alleged. Nobody wants a maladministration complaint against them for failing to properly consider a complaint about maladministration.
In his latest annual report for 2025-26, the Pensions Ombudsman reported another rise in complaint volumes (14%) against the previous year. That trajectory will no doubt steepen as AI-assisted submissions become routine.
Longer complaints require longer rebuttals; longer rebuttals lead to more complex determinations. The investigative and adjudication burden grows. Also, it is likely that the Financial Ombudsman Service will face the same dynamic across its broader jurisdiction. Neither body has unlimited capacity.
Using AI to resolve AI complaints
Of course, the same tools that generate complex complaints can also assist trustees, complaint-handlers, and the Ombudsman with triage, issue-spotting, and chronology drafting. AI can help identify the real dispute in a lengthy submission and flag genuine maladministration risks.
The Pensions Ombudsman recognises the opportunity to use technology and AI to help it “respond to the challenge of rising demand” for its services. It announced in October that, as part of its Digital Transformation Programme, it will use “automation and AI where appropriate to improve efficiency and capacity”.
Meanwhile, the Pensions Regulator has issued helpful guidance on the responsible use of AI in pension schemes. This even gives an example of an AI system that triages member complaints.
It has also issued a timely reminder that “the use of AI does not change existing duties and responsibilities as a trustee, including… fiduciary duties”. Trustees “remain responsible for making sure [the] scheme is governed and administered effectively and in line with legal and regulatory requirements, regardless of which technologies are being used”.

So where does this leave us? Is AI a friend or foe in the pensions space, or do we still not really know?
Well, AI and AI-generated complaints are here to stay. The pensions industry therefore needs to get better at separating the genuine grievance from the legal scaffolding built around it. That means resisting the instinct to answer every AI-generated argument with an equally lengthy rebuttal and focusing instead on the substance of what has actually happened.
The irony may be that the best answer to the challenges created by AI is more AI, but coupled with something technology cannot replace: judgment, proportionality, and common sense.
Charlotte Scholes and Hayley Goldstone are both partners at Pinsent Masons.











