Generative AI chatbots are proving capable of answering most pension member queries accurately and safely, with new research showing one scheme’s AI tool successfully resolved around 90 per cent of questions without human intervention.
The research was carried out by communications consultancy Quietroom. It includes an analysis of thousands of anonymised conversations between Ford UK employees and PensionsChat, a generative AI chatbot designed to support members of Ford Motor Company’s UK pension schemes.
The study suggests GenAI is already improving member outcomes by giving employees clear, practical and usable answers, while staying within the advice and guidance boundary. Rather than replacing existing support, the technology appears to be widening access to trusted information and reducing friction for members seeking help.
The main point of the research was to highlight the importance of high-quality data. It notes that PensionsChat’s performance is underpinned by clear, well-written and well-structured scheme documentation, built around the questions employees actually ask. The report argues that without this foundation, GenAI tools are unlikely to deliver reliable results.
The research also found that employees did not need support to interact with the chatbot; instead, asking questions in their own words and received accurate responses in seconds, suggesting GenAI tools are intuitive even for complex subjects such as pensions.
Quietroom says the tool’s effectiveness also reflects strong governance. Unlike open AI assistants, PensionsChat is grounded in Ford’s internal documents, with responses tested by pensions experts and supported by guardrails designed to manage risk around advice and guidance.
Most usage takes place during working hours, but a significant proportion of employees also access the chatbot before and after work and at weekends, highlighting the value of 24/7 availability.
