unthinkable
Insurers · AI for Business Operations

Your policyholders are meant to get the cover they need, and get paid when they claim, without chasing anyone.

For insurers, MGAs and insurance distributors

We can't change your underwriting appetite, your rates or what a policy covers. We help you remove everything else between policyholders and the cover and claim payment they are entitled to: the forms, the missing documents, the phone calls, and the not knowing.

Where policyholders have to chase today

35%of UK car insurance complaints the ombudsman decided went the customer's wayFinancial Ombudsman Service, 2025/26
38%of UK buildings insurance complaints the ombudsman decided went the customer's wayFinancial Ombudsman Service, 2025/26
10%of property and casualty claims are fraudulent, by Deloitte's estimate, so honest claims wait on checks tooDeloitte Insights, 2025
The policyholder's journey

The policyholder's journey from a quote to a paid claim, without chasing.

Seven moments from the first quote to a paid claim, and to a straight answer when something goes wrong. For each: what changes for the policyholder, what AI does (and what people still decide), and how we help you build it.

Today: a wait before every stepWith AI at work: straight through

1 · Getting a quote

A quote in one conversation, with what it covers and leaves out in plain words.

BUILD AI IN

An agent asks only the rating questions you don't already have answers to, returns the premium from your rating system and explains cover and exclusions from the policy form. A licensed agent advises on which cover to buy.

BUILD WITH AI

Built on your rating engine and policy forms, so prices and wording stay the ones you filed.

Software we've built for this step: Insurance marketplaces & operations →

2 · Buying the cover

Apply once, with what you already hold filled in and every document asked for upfront.

BUILD AI IN

What you already hold about the customer is pre-filled, underwriting questions come from your guidelines and uploaded documents are checked against the application. Underwriters decide anything outside the rules.

BUILD WITH AI

On your policy administration system and customer record; the policy is issued there, not by the agent.

Software we've built for this step: Insurance marketplaces & operations →

3 · Changes & renewal

Policyholders see why a renewal changed, and make routine changes in a message.

BUILD AI IN

Renewal changes are explained from the rating factors on record, and routine endorsements such as an address, a driver or a deductible are re-rated and confirmed by the customer. Anything outside the rules goes to a person.

BUILD WITH AI

Reads and writes through your policy system, so every change is a normal endorsement with its own record.

Software we've built for this step: Insurance marketplaces & operations →

4 · Reporting a claim

Report a claim in one conversation, from the scene, and know who has it straight away.

BUILD AI IN

The agent takes the first notice of loss, checks the policy was in force and what cover it carries, collects photos and details, and opens the claim with an adjuster. Anyone hurt or in danger goes straight to a person.

BUILD WITH AI

On your claims platform (Guidewire, Duck Creek or your own) through its APIs.

Software we've built for this step: Insurance marketplaces & operations →

5 · The assessment

Policyholders always know what their claim waits on, and who has it.

BUILD AI IN

Claim stage, adjuster and documents are read from the claims system, and entitlements such as rental cover from the policy. Exact checks run on every claim so clean ones move on and unusual ones go to an adjuster with the evidence. Adjusters decide coverage.

BUILD WITH AI

On your claims platform through its APIs, with fraud checks as rules your claims and SIU teams can read.

Software we've built for this step: Insurance marketplaces & operations →

6 · Getting paid

Claimants see how their payment is worked out, and when it will arrive.

BUILD AI IN

The payment is calculated by code from the approved amount, limits and deductible, and explained from the policy. An adjuster approves the amount and a person on your payments team releases it.

BUILD WITH AI

Connected to your claims platform and payment systems, with every payment logged.

Software we've built for this step: Insurance marketplaces & operations →

7 · Questions & complaints

A disputed decision is explained at once, and reaches a new reviewer the day it's raised.

BUILD AI IN

The agent explains the decision from the letter and the policy clause it cites, gathers the new evidence and opens a complaint with its deadline. A claims manager, not the original adjuster, reviews it.

BUILD WITH AI

Reads your claims, policy and complaints records; every step logged for market conduct examiners.

Software we've built for this step: Insurance marketplaces & operations →

What changes for each role (text version)

Getting a quote

  • What if a quote came with what it covers, in plain words? Today: People compare quotes without knowing what each one covers or leaves out. With an agent: Customers get a quote in one conversation, with cover and exclusions explained from the policy form. Rule, not model: Who is eligible and what the premium is come from underwriting rules and the rating system, never from the model.

Buying the cover

  • What if buying a policy didn't mean typing in what the insurer already knows? Today: Applicants fill in forms that ask for details the insurer already holds, then learn later what is missing. With an agent: Customers apply in one conversation, with what you hold filled in and every document asked for upfront. Rule, not model: Underwriting questions and the documents each product needs are underwriting guidelines in code.

Changes & renewal

  • What if a renewal notice told you why the price changed? Today: Renewal notices show a new premium without saying why it changed, so policyholders call. With an agent: Policyholders see what changed at renewal, from the rating factors on record, and make routine changes in a message. Rule, not model: Premiums, rating factors and which changes a policyholder can make alone come from the rating system and policy rules in code.

Reporting a claim

  • What if reporting a claim took one conversation, from the scene? Today: Policyholders report a claim by phone, send photos and papers separately, and wait to hear who has it. With an agent: Policyholders report a claim in one conversation, send photos from the scene and know who has it straight away. Rule, not model: Whether the policy was in force and what cover it carries come from the policy record by code.

The assessment

  • What if policyholders always knew what their claim was waiting on? Today: Deloitte estimates about 10% of property and casualty claims are fraudulent, so honest claims wait on checks too. With an agent: Policyholders get updates at each step, and claims that pass every check move straight on to the adjuster's approval. Rule, not model: Coverage limits and entitlements come from the policy record by code, not from the model. (Deloitte Insights, 2025)

Getting paid

  • What if claimants knew the payment, and why, before it arrived? Today: Claimants often learn the payment amount and the deductible only when the money arrives. With an agent: Claimants see how the payment is worked out from the approved amount and the policy, and when it will arrive. Rule, not model: The payment is calculated by code from the approved amount, limits and deductible; the model never works it out.

Questions & complaints

  • What if a disputed claim reached a new reviewer the day you asked? Today: The UK ombudsman upheld 38% of buildings insurance complaints it decided in 2025/26. With an agent: Policyholders hear the decision explained from the policy, and a disputed decision goes to a new reviewer with the evidence ready. Rule, not model: Policy wording is quoted from the issued policy, and complaint deadlines are set in code from the rules that apply. (Financial Ombudsman Service annual complaints data, 2025/26 (UK))

Try it with your own case →

The real problem

Why policyholders have to chase for cover or a claim.

Your systems hold the answers. People still have to go and find them.

Rating, policy administration, billing, claims and complaints each do their job, but they don't talk to each other. So staff become the glue: they re-key applications, chase photos and estimates, and look up claim status for every call. And policyholders chase, first by phone and then by complaint.

Honest claims wait behind suspicious ones

Deloitte estimates about 10% of property and casualty claims are fraudulent, so every claim gets checked by hand before it is paid (Deloitte Insights, 2025).

A no without its reasons becomes a complaint

The UK ombudsman decided 35% of car insurance and 38% of buildings insurance complaints in the customer's favour in 2025/26 (Financial Ombudsman Service).

Fraud costs land on every claim

Deloitte estimates property and casualty fraud costs about $122 billion a year in the US. Checks that run as code let clean claims through instead of holding every claim for a manual look (Deloitte Insights, 2025).

Which waits software can take out, and which it can't.

Getting a quote
Buying the cover
Changes & renewal
Reporting a claim
The assessment
Getting paid
Questions & complaints
The whole journey
Wait made by process: handoffs, approvals, lookupsWith AIThe work itself: capacity

Lengths are illustrative, not measurements. In each step, software can take out the wait that process makes: handoffs, approvals and lookups. It can't add capacity, so the work itself stays.

How we help

How we help.

BUILD AI IN

Agents across the policyholder's journey from a quote to a paid claim, each with one job: they answer, collect, check and explain. Underwriters, adjusters and licensed agents make every underwriting, coverage, payment and advice decision. Rules that must be exact (eligibility, premiums, coverage limits, payment amounts, claim deadlines) run as code, not as a model's guess.

BUILD WITH AI

On the systems you already run (your rating engine and policy administration system, Guidewire or Duck Creek for claims, your document and complaints systems), one moment at a time. Your team owns the code and the data.

The whole insurance operation, by the moment it affects.

Getting a quote
  • Quote and rating
  • Product eligibility
  • Agent and broker channels
  • Website and contact centre
Buying the cover
  • Application intake
  • Underwriting questions and referrals
  • Identity checks (KYC)
  • Policy issue and documents
Changes & renewal
  • Endorsements
  • Renewal notices
  • Billing and payments
  • Document archive
Reporting a claim
  • First notice of loss
  • Coverage checks
  • Claim photos and documents
  • Adjuster assignment
The assessment
  • Claims triage
  • Fraud checks
  • Repairs and rentals
  • Estimates and inspections
Getting paid
  • Settlement calculation
  • Payment release
  • Recovery from other insurers
Questions & complaints
  • Policy questions
  • Claim decision letters
  • Complaints
  • Market conduct exams

Building a product for insurers? Insurtech →

Safety

Is it safe?

BUILD AI IN
No underwriting or claim decision by a bot
Agents never bind cover, set a premium, decide coverage or approve a payment. Underwriters and adjusters decide, and the agent never hints at a different outcome.
Wording from the policy, not the model
Premiums, limits, deductibles and clauses come from your systems and issued policies by code; the model only explains them in plain words.
Every action logged
What the agent read, the rule that applied and who approved, replayable for a market conduct examiner.
BUILD WITH AI
Policyholder data stays where your rules require
Data stays in your approved environment, each agent has a written access list, every read is logged, and nothing is used to train models (GLBA).
AI governed the way regulators ask
Each AI step is documented, tested before release and logged, so you can show how it is governed, in the spirit of the NAIC model bulletin on insurers' use of AI.
Sandbox first
Built and tested in your vendors' test environments before any live data, with your security and compliance review planned from the start.
Proof

Results with insurers.

We have no claims-agent case study to show yet, so we show the insurance work we have done: an insurtech app, and the policy records and documents behind it.

ACKO, India (insurtech)90+companies onboarded within months (as reported)

Employees of client companies see all their cover in one place, track claims and enrolment, and are told which documents are missing. The corporate health insurance app had been slow and couldn't handle several policies per person; it was rebuilt.

Read the case study →
A leading Indian general insurer60TBmigrated with no data loss (as reported)

Customers upload policy documents and vehicle photos to a system that keeps up, with ID and registration numbers indexed as they arrive. Its document system had dropped support for DB2, and 60 TB of live data had to move without stopping the business; it was moved to Oracle 19c.

Read the case study →
Pramerica Life Insurance, India98%accuracy across 500,000+ documents (as reported)

National ID numbers are masked in every stored customer document, as a new government rule required, ahead of the deadline. About 500,000 hand-scanned documents held the full number in different places; bots found and masked each one.

Read the case study →

Which step would you stop policyholders chasing first? Name the moment where your policyholders call most. After a short discovery you get a plan with dates for the first release, measured against those calls.

Tap one: Ask Unthinkable starts your plan from there. Or say it in your own words ↓

How to choose

How to choose who builds AI for insurers.

Four things to check before you choose who builds it, and how we meet them.

Has worked on policies and claims before

Group health cover and claims tracking for ACKO, a 60 TB move of an Indian general insurer's policy-document system, and ID masking across 500,000+ stored documents for Pramerica Life Insurance (as reported).

Often elsewhereNew to insurance, or know it only from a requirements document.

Starts from your policyholder's journey

We follow the seven moments from a quote to a paid claim and a complaint. We also build insurtech products, so we know what the software on each side needs.

Often elsewhereConsultants know insurance but don't build; software companies build to the spec.

Agents that do the work

At the assessment, for example, the agent tells the policyholder what the claim is waiting on and books the rental the policy includes; adjusters still decide.

Often elsewhereA chatbot on the website, or a dashboard someone still has to act on.

Exact where a regulator looks

Eligibility, premiums, coverage limits, payment amounts and claim deadlines run as code, and decision letters quote the issued policy, so every step can be shown to a market conduct examiner.

Often elsewhereA quick pilot that never passes compliance review.

The full comparison: us, other software companies, in-house teams and consultants →

Ways to work

Start with one moment in your policy or claims operation. Measure it. Grow from there.

Time & materials

A senior AI team billed for time, within an agreed monthly budget. Even on time and materials the work is tied to your business objective: every plan starts from it, and every report shows the number next to the hours.

One fixed outcome

One moment, one agreed measure, a fixed price.

Dedicated team

A complete senior AI team, monthly, working with your IT, claims, underwriting and compliance leads.

Linked to the result

Where the gain can be measured: a reduced implementation fee plus a percentage of the expected gain (fewer status calls and complaints, fewer claims stuck waiting on documents).

How we work and price →

Straight answers

Questions insurers ask.

How should an insurer choose a partner to build AI agents?

Pick the moment where your policyholders chase you most, such as claim status, a renewal increase or a declined claim, and ask each candidate how they would take the chasing out of it inside your rating, policy and claims systems. Listen for three things: underwriting, coverage and payment decisions stay with your underwriters and adjusters; premiums, limits and claim deadlines run as code your compliance team can read; and every action is logged so you can show a market conduct examiner what happened. Ask how they test in your vendors' sandboxes before live data, and confirm you will own the code and data. Unthinkable works this way and also builds insurtech products, so it knows both sides. Its insurance work includes ACKO's group health insurance app and ID masking across 500,000+ stored documents for Pramerica Life Insurance (as reported).

Can AI decide insurance claims for us?

It shouldn't decide them alone, and ours don't. Agents take the first notice of loss, check the policy was in force, collect photos and estimates, run exact checks in code and keep the policyholder told. A named adjuster decides coverage and approves every payment, and a person on your payments team releases it. Decision letters quote the issued policy, and the adjuster approves each one before it is sent.

Can AI help detect insurance fraud without slowing honest claims?

Yes, by separating the two. Deloitte estimates about 1 in 10 property and casualty claims are fraudulent. Agents run exact checks in code, such as policy dates, duplicates and linked parties, and read photos and documents for inconsistencies. Clean claims are fast-tracked for approval, and unusual ones go to an adjuster with the evidence. A signal is never an accusation; only people refer a claim or deny it.

How can AI reduce policyholder complaints?

Mostly by explaining before policyholders have to chase. The UK ombudsman decided 35% of car insurance and 38% of buildings insurance complaints in the customer's favour in 2025/26. Agents tell policyholders what their claim is waiting on, explain renewals and decisions from the policy, and send a disputed decision to a claims manager with the evidence gathered. Replies to regulators stay with your staff.

Can an AI agent explain quotes without giving insurance advice?

Yes, if the line is drawn in the design. The agent asks rating questions, returns the premium from your rating system and explains what the policy form covers and leaves out. Recommending which cover a customer should buy stays with a licensed agent, and the agent hands the conversation over when a customer asks for advice. Which questions need a licence depends on your states and products.

Can you integrate AI agents with Guidewire, Duck Creek or our own policy and claims systems?

Yes. We connect through the APIs your rating, policy administration and claims platforms support, such as Guidewire or Duck Creek, and build in a test environment first. Each agent reads and writes only what its access list allows, and every write is logged. Our team has also moved an insurer's 60 TB document system from DB2 to Oracle without data loss (as reported). Exact scope depends on your versions and vendor agreements.

What does it cost to bring AI agents into an insurer?

One moment is usually a fixed piece of work; several moments are a monthly team. Where the gain is measurable, such as fewer status calls, complaints or claims waiting on documents, part of the fee can be linked to it. We give you a written estimate after a short discovery.

Is our policyholders' data safe with AI agents?

Yes, when the agents run inside your controls. Policyholder data stays in your approved environment and is never used to train models, each agent reads and writes only what its access list allows, and every action is logged for your examiners. Unthinkable designs to the controls your auditors already use, such as GLBA and GDPR requirements; for Pramerica Life Insurance, our team masked national ID numbers in 500,000+ stored documents (as reported).

Ask Unthinkable

Where do your policyholders chase most today?

Tell us where policyholders chase you today, for a quote, a renewal or a claim. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.

Send us a note instead

Required

Thanks. A senior AI engineer will reply to you.
u
Ask UnthinkableClarity first, then the right team
Your conversation

An AI agent. This prototype uses prepared answers and sends nothing.