The use of AI tools (eg ChatGPT or Claude) by pension scheme members is no longer hypothetical. It is already reshaping how some complaints are drafted, framed and pursued.
From generating structured complaint letters to assisting with legal research and tone, AI tools give members access to resources that can produce complaints which are more detailed and, in some cases, more legally sophisticated than ever before.
While this shift can improve accessibility to a scheme’s IDRP, it also presents new challenges for trustees and administrators responsible for handling complaints and anticipating potential escalation to TPO.
Common features of AI-assisted complaints
When reviewing a complaint, there are a number of features that may indicate AI has been used to assist with drafting:
- Lengthy complaints – members can now simply use prompts such as “add more detail” and be provided with a five-page IDRP complaint, even where the underlying issue could be summarised in a few paragraphs.
- Structured headings and bullet points – most AI models favour a highly structured format, often using headings, sub-headings, bullet points and bold text to present arguments.
- Adversarial tone – complaints may increasingly rely on references to case law and statutes to support arguments, sometimes resulting in a more confrontational tone.
- Confident assertions with limited supporting evidence – AI-assisted complaints can present arguments with a high degree of confidence, even where the legal authority or factual basis is weak, incomplete or inaccurate.
- Repetition of the same point in different ways – AI tools often expand complaints by restating the same argument using different wording. This can create the impression that multiple issues are being raised when, in reality, there may only be one or two core concerns.
Key risks and top tips for trustees
- Don’t panic – the presence of AI-assisted content does not necessarily mean a complaint is stronger or more likely to succeed.
- Focus on the substance – a lengthy complaint does not always equate to more legal arguments. Before drafting a response, identify the key issues that are genuinely in dispute, leaving aside any repetition.
- Always check the law – AI models can “hallucinate” legal authorities (ie they can cite incorrect or made-up authorities that often appear plausible), misquote cases or misinterpret judgments and TPO determinations. Always verify any legal references before responding.
- Use plain English – although a complaint may contain complex legal terminology, members are increasingly likely to paste an IDRP response into an AI tool and ask, “What does this mean?”. Clear, straightforward language can reduce the risk of misunderstanding and further disputes.
- Respond to the issues, not the volume – trustees should ensure that every substantive complaint point is addressed, but they should resist the temptation to respond separately to every repetition or variation of the same argument. A structured response that deals with the underlying issues is often more effective.
Looking ahead
The use of AI in member complaints is likely to continue to increase. While it may improve accessibility and engagement for members, it also requires trustees to adapt their approach.
In addition to changing how complaints are drafted, AI may also increase the overall volume of complaints. Members who might previously have been deterred by the time, effort or confidence required to prepare a complaint can now prepare a detailed submission in a matter of minutes. This is a practical challenge that some trustees and administrators may need to consider, particularly where complaint-handling resources are already stretched.
Importantly, AI does not change the underlying legal framework. Trustees’ duties remain the same: to act in accordance with the scheme rules, the law and the principles of good administration.
By focusing on substance over style and adopting a structured review process, trustees can respond to complaints effectively, regardless of how they are drafted.
