The most AI-intensive part of UK finance. And the most exposed.
Insurers use AI in pricing, claims, underwriting and service more than any other financial sector, and much of it sits inside supplier systems. The FCA has said it is reviewing how AI is used in underwriting, claims and customer service.
of UK insurers use AI
Bank of England and FCA survey, 2024use cases expected per firm within three years, up from nine
Bank of England and FCA survey, 2024rate data protection as a large constraint on AI
Bank of England and FCA survey, 2024Outcomes for customers, and a named person who answers for them.
The FCA's 2026 priorities for insurance include a review of how firms use AI in underwriting, claims and customer service, while remaining supportive of innovation. It expects firms to monitor customer outcomes and manage the risks, under the Consumer Duty and the Senior Managers regime. It is also widening its work on oversight of outsourced claims handling, with findings expected in early 2027. Health information used in claims is special-category data under UK GDPR, and the Data (Use and Access) Act 2025 does not relax those protections. The PRA's model risk statement SS1/23 is written for banks; the PRA has said it may extend model risk expectations to insurers, so check current scope.
Summarised from FCA 2026 insurance priorities (as reported by law firms) and PRA publications. Not legal advice.
Six problems that keep coming back.
Drawn from the FCA's stated priorities and public surveys of the sector.
Triage, document reading and fraud flags now shape who is paid and how fast. If a customer or the Ombudsman asks why, you need the reasoning and the person who checked it.
Pricing models change often and draw on many data sources. Showing that they produce fair outcomes for each customer group depends on records that most teams do not keep.
Third parties handle claims for you, and their systems use AI you did not choose. The FCA is looking harder at how insurers oversee those chains.
Medical reports and injury descriptions are special-category data. Pasted into a public tool, they become a breach, not just a policy lapse.
Bereavement, ill health and financial stress show up in claims. AI-drafted letters and chat must recognise them and hand over to a person.
A supplier updates a model and claims decisions shift. Without sandbox testing and a change record, no one can say when, or why.
A claims dispute is decided on the record you can produce.
When a declined claim becomes a complaint, or a pricing question reaches a supervisor, the insurer that can show the data used, the rule applied and the person who signed off holds the stronger position. The insurer that cannot spends weeks reconstructing events from supplier logs, and carries the reputational cost of a customer who was told 'the system decided'.
Supplier AI is your AI.
The Consumer Duty and the Senior Managers regime do not stop at the contract boundary. An AI feature inside a policy administration or claims platform needs an owner, a data classification and a review, just as a model you built would. The fix is a register that includes supplier features, plus rules that apply wherever you route requests.
- Claims triage and document extraction
- Underwriting support and summarisation
- Customer service chat and voice
- Fraud indicators
- Special-category (health) data in prompts
- Pricing and claims models without a documented human check
- Supplier AI inside policy administration systems
GenWA registers supplier and in-house AI, redacts health and financial data on connected tools, and records named exclusions where a person must decide, such as declining a claim or setting a price. Sandbox testing before a model switch is part of the design.
Four agents, trained for this sector.
The rules, as controls.
- FCA Consumer Duty
- FCA insurance conduct rules (ICOBS)
- Senior Managers and Certification Regime
- UK GDPR (special-category data)
- Payment card security standard (PCI DSS)
Built around how insurers and managing agents are organised.
Carriers, managing general agents and brokers carry different duties. We shape the route to yours.
An independent AI audit that includes supplier-embedded AI, with first findings in 48 hours and a 90-day plan for your risk committee.
Start the health checkAdvice: policy and controlsAn AI policy and control set that fits your delegated authority arrangements, so a managing general agent and its carrier share one view.
See advisoryProduct: control toolkitRegister, Guard rules, log and Evidence Packs, hosted in your AWS or Azure account, or GenWA AWS as agreed.
Find the right productProduct: agents at workClaims document extraction, complaints triage and adviser assist with named owners.
Meet the teamIncubator: one claims or service functionWe define, build and run one controlled function in twelve weeks. You own the intellectual property, prompts and record.
See the incubatorAccelerator: for insurtech foundersBy application only, ten places a year. Senior AI leadership time, advisory and platform access, paid for by equity, project, retainer or licence.
See the acceleratorWhat Guard and Evidence typically hold for this sector.
Control-mapped, not certified. We prepare the evidence; regulators and accredited bodies decide.
Good outcomes for retail customers, mapped to claims, pricing and vulnerability controls.
FCA rules on how insurers sell and service policies, mapped to customer communications.
Named accountability for each AI use.
Model risk principles, mapped as good practice.
Special-category data handling, redaction and automated-decision safeguards.
AI management and service-controls standards. Readiness evidence; accredited bodies certify.
Three role holders, three different problems.
Needs to explain why a claim was triaged, paid or declined. GenWA records what the AI read, what it suggested and who decided, so complaints are answered from the record.
Owns pricing and risk selection. GenWA names where AI must not decide, tests model changes in a sandbox first, and keeps the change history.
Must show the board and the supervisor which AI is in use, including inside suppliers. GenWA gives one register, one rule set and evidence packs on request.
From this sector.
What does GenWA provide for insurers using AI in claims, underwriting and pricing?
A register that includes supplier AI, redaction of health and financial data on connected tools, rules with named exclusions where a person must decide, and sandbox testing before model changes. The record shows who accepted each output. People decide on claims and pricing.
We use a claims platform with AI built in. Is that covered?
Within the scope we agree. The register records supplier AI features with an owner, a purpose and a data classification. Controls such as redaction apply to connected, routed tools, so coverage depends on the integrations set out in the audit.
How does GenWA treat health and special-category data?
Data Protection removes personal and special-category data before approved external model calls on connected tools. Hosting is your AWS or Azure account, or GenWA AWS as agreed. Detailed coverage for health data is confirmed in the security review.
Which regulators and rules apply to AI in UK insurance?
The FCA and the PRA, plus UK GDPR. The FCA is applying the Consumer Duty and the Senior Managers regime to AI. The PRA's model risk statement SS1/23 is aimed at banks, so check whether and how it reaches your firm. We map controls to the rules you name. Not legal advice.
How many UK insurers use AI?
The Bank of England and FCA survey of 2024 found 95% of UK insurers use AI, and expected 21 use cases per firm within three years, up from nine. It is historical and covers financial services only, so use it as context, not a forecast.
What happens when a supplier changes a model?
Change Control keeps a catalogue of the models in use and briefs you when one changes. Sandbox testing before a switch is part of the design. Tested switches and rollback are being built next and are not live.
How does an insurer start?
Take the AI health check and name the uses you care about, such as claims or pricing. First findings arrive within 48 hours and the audit with a 90-day plan within ten working days. The audit fee is credited in full against year one of a platform plan.
Which of your claims decisions has AI touched?
First findings within 48 hours of the health check. Audit fee credited in full against year one of a platform plan.