By
Stuart Kerr, Technology Correspondent,
LiveAIWire

Ninety-five per cent of UK insurance firms now use artificial
intelligence somewhere in their business, the highest adoption rate of any financial
sector, according to the Bank of England and Financial Conduct Authority’s
2024
survey of AI in UK financial services. That places insurers ahead
of banks and investment managers. It matters now because the same technology
deciding how fast your claim is paid is also deciding what you are charged,
and increasingly whether you can get cover at all.

The Premium You Pay Is Now Set by a Machine

Pricing and underwriting, once the work of human actuaries poring
over tables, are now largely machine tasks. The European insurance regulator
EIOPA found that 50 per cent of non-life insurers and 24 per cent of life
insurers across Europe were already using AI across the value chain,
including pricing and underwriting, in its 2024
digitalisation review. The appeal for insurers is obvious. Models
trained on thousands of data points can price a risk in seconds and spot
patterns a person would never see.

For many customers that means faster quotes and, in low-risk
cases, lower premiums. A careful driver with a clean record and a telematics
box may pay less than they would have under a blunt human estimate. The
granularity cuts both ways, though, and the closer a model looks, the more it
can separate one customer from another.

Faster Claims Are the Industry’s Easiest Win

The clearest benefit shows up after something goes wrong. AI
systems now triage claims, read damage photographs, flag likely fraud and
settle simple cases without a human ever opening the file. EIOPA’s review
identified claims management and fraud detection as two of the most common
uses across European insurers, alongside pricing.

That speed is real and it favours the customer. A motor claim that
once took weeks can clear in days when software handles the routine paperwork
and routes only the awkward cases to a person. Fraud detection, meanwhile,
protects the honest majority, because every fraudulent payout is eventually
recovered through everyone else’s premiums. This is the part of the story
insurers are happy to tell, and on the evidence they are right
to.

The gains are easiest to see in motor and home cover. An insurer
using image recognition to assess a dented bumper from a few phone
photographs can issue a settlement before a human assessor would have booked
a visit, and that convenience is genuine. The same models that read those
photographs also scan for the tell-tale inconsistencies of a staged claim,
quietly screening out fraud that would otherwise be absorbed by every honest
customer through higher renewal prices. Where AI sticks to this work, sorting
the routine from the suspect, the case for it is strong and the customer
mostly benefits.

The People Algorithms Quietly Price Out

The harder story sits in who gets left behind. EIOPA warned in its
2024 review that AI driven pricing can lead to excessive standardisation and
a limited consideration of a customer’s specific circumstances. In plain
terms, a model optimised for the average can misread anyone who is not
average, and the people most often misread are those already on the
margins.

When a system prices on hundreds of correlated signals, it can
rebuild a picture of someone’s health, income or neighbourhood without ever
being told those things directly. The result can be higher premiums or quiet
exclusion for higher-risk and vulnerable customers, a concern EIOPA has
raised repeatedly as it presses supervisors to watch for unfair outcomes.
None of this requires malice. It is simply what happens when accuracy becomes
the only goal and fairness is left to look after itself. The same tension
surfaces in our coverage of AI
and the great insurance gamble, where finer prediction and fairer
treatment pull against each other.

The Black Box Problem Insurers Cannot Yet
Explain

Speed and accuracy come with a cost the industry is still
wrestling with: many insurers cannot fully explain how their own systems
reach a decision. The Bank of England and FCA survey found that a large share
of firms hold only a partial understanding of the AI they deploy, and that
foundation models, the complex systems behind generative AI, already account
for around 17 per cent of all AI use cases in UK financial services. When a
model is opaque even to the company running it, a customer trying to contest
a price or a refusal faces a steeper climb still.

This is why supervisors keep returning to transparency. A premium
you cannot interrogate is a premium you cannot challenge, and an insurer that
cannot show its working cannot easily prove it treated you fairly. The same
survey recorded that firms themselves see cybersecurity as the single
greatest risk attached to their AI, a reminder that the systems now holding
your most sensitive financial and medical data are also a target for the
people who would steal it. The more of the insurance relationship that
disappears into these models, the more the question of who can actually see
inside them comes to matter.

What This Means for You

For an ordinary policyholder, three practical habits now matter
more than they used to. Shop around harder, because two insurers running
different models can return startlingly different prices for the identical
risk, and loyalty is rarely rewarded. Ask what data a quote is based on,
since under UK data protection law you are entitled to know, and you can
challenge a decision made solely by automated means. And read the detail on
telematics or app-based policies before signing, because the discount on
offer is paid for with a continuous stream of data about how you drive, live
and move.

Consider a self-employed courier refused affordable motor cover
because a model reads gig-economy mileage as elevated risk. The fix is rarely
to argue with the algorithm. It is to find an insurer whose model weighs that
profile differently, which is why comparison and a willingness to switch are
now a consumer’s strongest tools. The wider shift from human to automated
decision-making across daily life runs through our look at the
automation divide.

The Next Five Years Will Decide Who Insurance
Serves

Regulators are no longer watching from the sidelines. Under the EU
AI Act, AI systems used for risk assessment and pricing in life and health
insurance are now classed as high-risk and face stricter requirements, and
EIOPA has issued guidance pushing insurers toward transparency and
accountability. Britain’s Bank of England and FCA are running their own
monitoring through repeated surveys and a dedicated AI consortium. The
question for the rest of this decade is whether that oversight keeps pace
with the models, or trails a step behind them. The technology that makes
insurance faster and cheaper for most people is the same technology that can
shut the door on a minority, and which of those outcomes wins will be settled
less by code than by the rules we choose to write around it. Britain has so
far favoured monitoring over hard legislation, leaving its regulators to lean
on surveys and supervisory pressure rather than statute, while Europe has
reached for binding law. Which approach better protects the customer who
finds themselves on the wrong side of a model is a question the next few
years will answer in public. For the deeper mechanics of how insurers turn
data into prices, see our guide to AI
in insurance, premiums and predictions.

About the Author

Stuart Kerr is Technology Correspondent at LiveAIWire, covering
artificial intelligence, cybersecurity, and the social impact of emerging
technology. He publishes daily at LiveAIWire.com.