AI for insurersfrom first submission to fair settlement.
From Lloyd's underwriters to personal lines: submissions arrive as emails and PDFs, claims take weeks, and core systems are decades old, while regulators ask you to prove every outcome is fair. We build AI for insurers that works within your wordings and your rules, inside your own walls, and records why every decision was made.
Follow a claim through ↓
Duncan Anderson, our co-founder, was Chief Architect at Legal & General, responsible for overseeing new core administration systems and the digitisation of the business. He knows what it takes to change an insurer's systems without stopping the business. Duncan on LinkedIn ↗
From the inside
Where insurancegets stuck.
Familiar problems to anyone who has run an insurer's operations or its technology. AI can fix them, if it's built to work within the rules.
Claims take too long, and the regulator has noticed
The FCA's 2024 review of life insurers found term claims took an average of 53 to 122 days from start to finish. Its business interruption review found handlers couldn't always find the right policy wording, which added delay.
Submissions arrive as emails and PDFs
Broker submissions, schedules and loss runs come in every format. Much of the data is retyped or never captured at all, and underwriters spend their time on admin rather than on risk.
Core systems nobody wants to touch
Life and pensions books run on platforms decades old, with key policy data in scanned documents. Migrations are so risky that many books simply run off on legacy systems while the cost per policy climbs.
Every outcome has to be fair, and provably so
Consumer Duty asks insurers to evidence fair value and good outcomes, in claims above all. Mapping FCA and PRA obligations to policies and controls is often a consultancy project, repeated with every change.
Who it's for
Every kind of insurerhas different problems.
A personal lines insurer, a specialty underwriter and a life and pensions provider run very differently. We start from how your business works.
Personal lines
Claims, complaints and renewals at volume, with fair value and good outcomes to prove under Consumer Duty.
Commercial and specialty
Submissions in every format, and underwriters who should be pricing risk rather than retyping schedules.
Life and pensions
Long-lived policies on old systems, bereavement claims that need care, and back books that cost more every year.
MGAs and brokers
Lean teams handling binders, bordereaux and broker emails, where every hour of admin comes out of margin.
The life of a claim
Every claim, start to finish,with every step on the record.
Claims are where an insurer keeps its promise. This is how the system we'd build handles one, and which part of it does the work at each step.
01 · First notice
Captured in one go
Phone, web or email: an agent takes the details, photos and documents, checks the policy is live, and opens the case.
ByPharos
02 · Cover
The wording that applies
The exact policy wording, schedule and endorsements for this customer, so the handler and the agent read the same clause.
03 · Assessment
Evidence gathered, anomalies flagged
Agents gather reports, estimates and claims history, and flag anything unusual for a person to look at. Vulnerable customers are spotted early.
ByPharos
04 · Decision
Rules decide, people approve
Coverage rules drawn from your wordings and claims policy decide; a handler approves anything that matters. Same claim, same answer.
05 · Settlement
A fair outcome you can evidence
Every decision recorded against the clause and the rule behind it, so you can show good outcomes and fair value under Consumer Duty.
Where it helps
Where AI paysfor an insurer.
Underwriting submissions
For a specialty underwriter, an MGA or a personal lines team, agents read broker emails, slips, schedules and loss runs in whatever format they arrive, extract the risk data, check it against appetite and referral rules, and hand the underwriter a submission ready to price, with the gaps and the questions for the broker already flagged. As the market moves to structured, digital placement, the same agents keep your data in step.
Worth: Underwriters spend their time on risk, not retyping, and quote while the broker is still waiting.
Claims, from first notice to settlement
Agents capture the claim, check cover against the right wording, gather the evidence and draft the decision. Handlers approve, and spend their time on the claims that need judgement.
Worth: Faster, fairer settlements, fewer complaints, and less leakage.
Wordings your systems understand
Turn policy wordings and claims policy into rules, so you can see which clause applies to a claim, and which rules a wording change touches.
Worth: Consistent decisions, and no hunting for the right wording.
Regulatory obligations and fair value
Whereas maps FCA and PRA obligations to your policies and controls and shows the gaps, including the evidence Consumer Duty asks for.
Worth: Less consultancy spend, and gaps found before a regulator finds them.
Back books and core administration
Agents read scanned documents and old system records to clean policy data before a migration, and Pharos runs servicing work around the legacy platform.
Worth: Cheaper servicing, and migrations with fewer surprises.
Proof
Built for regulatedfinancial services.
Whereas: prove every obligation is met
Our compliance product maps risks to obligations, policies, rules, controls and data, so you can see what's covered, what isn't, and what a change would touch.
See Whereas ↗Falcon: the same approach, in banking
Our customer due diligence platform for large banks: hundreds of pages of policy turned into executable rules, agents that gather the evidence, and a case file for every decision.
What we learned building it →
The question boards ask
When AI is in every claim, who are you depending on?
Once AI reads submissions, prices risks and handles claims, the provider behind it becomes part of how you serve every customer. Prices, models and terms can all change, and an outage lands on your claimants.
The PRA expects insurers to have an exit plan for every material outsourcing, including a stressed exit, and Lloyd's managing agents face the same scrutiny. We design that exit in from day one, so no AI provider is one you can't leave.
Switch models without rebuilding
Change AI provider, or bring models in-house, without starting again. We choose the right model for each job, and resell none of them.
Policyholder data stays with you
Systems run in your own cloud and jurisdiction, and sensitive work can use open-weight models that never leave your walls.
You own the code
No licence to us and no platform to rent. Your team can run and change the system without us.
What happens next
From one processto a system you own.
A one-day workshop
Pick the process worth fixing first: a claims journey, submission intake, a wording you can't trace, or a back book.
A working prototype in two weeks
On sample or anonymised data, for your team, and your risk and compliance functions, to challenge.
Build it in our AI factory
Agents plan, build, test and security-scan; our engineers approve the plan and the result. Built to pass your security and model risk reviews.
Into production
On your own infrastructure, alongside your core systems, with the evidence your risk and compliance teams will ask for. The code is yours.
Running claims, underwriting or a back book?Let's talk.
Tell us the process you'd most like to fix, and we'll show you how it would run within your wordings and your rules.
Talk to Duncan
Co-founder, Barnacle Labs · Former Chief Architect, Legal & General