Only 30% of UK adults hold life insurance or income protection, despite 84% holding some form of general insurance, according to the Mills Review, the FCA-commissioned study into AI's impact on financial services published this month. One of the review's proposals is a personal AI agent capable of identifying protection gaps and prompting consumers to consider cover at the right life moment.
The figures track closely with the regulator's own separate findings. Interim results from the FCA's competition review of pure protection products, published in January, found 58% of UK adults hold no life, critical illness or income protection cover at all. Graeme Reynolds, the FCA's director of competition, said at the time that the regulator would "work with industry to reduce this gap."
That same review found the gap exists mainly because protection is "sold, not bought", not because of price: consumers are rarely prompted to think about cover and, when they are, the products involve difficult judgement calls about health disclosure, exclusions and what counts as incapacity.
Protection has long been a difficult product to sell because consumers rarely set out looking for it in the way they might shop for home or motor insurance.
"Nobody wakes up with a stonking urge to buy protection," said Justin Harper, chief marketing officer at LifeSearch. "It's not a natural thing, we think, 'Oh, I must have that. That's on the shopping list.'"
Harper said AI was beginning to change that behaviour as consumers increasingly turned to AI tools to answer highly personal insurance questions before ever speaking to an adviser. Rather than searching for the cheapest life insurance policy, people were asking AI increasingly detailed questions based on their own circumstances, such as their employment status or medical history. The average search query had grown to around 26 words as consumers sought more tailored guidance.
That is exactly the opportunity identified in the Mills Review. By making protection more discoverable and surfacing relevant products at the right life moment, AI could encourage more consumers to consider cover who otherwise never would.
The review also sounds a note of caution. In its summary of responses to the FCA's engagement paper, consumer advisers and advocacy groups warned that people "may over-trust AI outputs, particularly where they are presented confidently, and may struggle to identify mistakes or inaccurate information", creating "clear risks of detriment" wherever people rely on AI "without independent verification."
Harper said advisers were already beginning to see those risks in practice.
"AI is allowing customers to be more informed than they ever were before. So when they come to us they have a better picture, a better experience, but also at the same time more likely to be misinformed than they ever were before because they may have plugged in the wrong thing and they trust ChatGPT."
He said advisers were increasingly speaking to customers who had already asked AI tools what type of protection they needed, creating a different starting point for conversations. While AI could help educate consumers and point them towards products they may wish to consider, it remained far less capable of understanding the personal context that underpins protection advice.
Harper said AI could play an important role in educating consumers and helping them understand the types of protection they should consider before speaking to an adviser. However, he argued the technology still struggled to account for the personal circumstances that underpin protection advice, from an individual's medical history to their wider family situation and long-term goals.
AI clearly has a role to play. Identifying who has a protection gap, prompting them to think about cover and helping them understand their options are all use cases that advisers and insurers are already embracing.
The harder question the Mills Review raises, and doesn't fully answer, is what happens at the next step: when an AI agent moves from flagging a need to recommending or arranging cover. For all AI's ability to educate and prompt consumers, that remains a decision that goes well beyond information alone. As Harper put it: "It provides information but doesn't replace judgement."
