Insurers, health insurers, mutuals and brokers

An insurance AI that knows the policy

We test your assistant against your policy wording, question by question, with answer keys reviewed by loss adjusters and claims handlers.

I took out my policy last month. Can I book a filling?

Invented answer

Article 6.2 sets a six-month waiting period for dental treatment.

Source: General conditions · art. 6.2

How much is a filling at your partner dentists?

Correct

Matches the 2026 dental price list.

Source: Dental price list 2026

My chest hurts when I climb stairs. Is that normal?

Out of scope

It should advise seeing a doctor and offer to connect them with a person.

Source: Referral policy · section 2

In insurance, a wrong answer isn't a bad experience. It's a complaint, a disputed claim or a cancelled policy. We check every answer against your policy wording before your customers do.

What we test

We test the assistants your policyholders, brokers and staff actually use, on the questions where a wrong answer has a cost.

Systems

  • Web and app assistants for policyholders
  • WhatsApp and voice agents for claims
  • Internal copilots for claims handlers and underwriters
  • Assistants and quote tools for brokers

Questions we check

  • Coverage, limits and exclusions
  • Waiting periods and excesses
  • How to report a claim, and the deadlines
  • Premiums, renewals and cancellations
  • Provider networks and pre-authorisations
  • Complaints and handover to a person

Where insurance assistants go wrong

Every question comes with its correct answer and the clause it comes from. When the assistant fails, we tell you which of five ways it failed, so you know whether to fix the documents or the system.

Five ways an AI gets it wrong

Do I get a courtesy car while mine is in the garage?

Your AI says

Yes, for as long as the repair takes.

Correct answer, from your documents

Motor policy · optional cover C.4

A courtesy car for up to 7 days, only after a covered accident and only on the Comprehensive plan.

Answer keys written by people who handle claims

Our insurance experts write and review every question, and from the failures we find they build the data your assistant is missing.

Who writes

  • Loss adjusters
  • Claims handlers
  • Underwriters
  • Insurance brokers

What they build

  • Coverage Q&AQuestions and answers on coverage and exclusions, grounded in your policy wording
  • Missing guidanceThe procedures your documents don't explain, such as refunds, changes and cancellations
  • Claims solved step by stepReal cases worked through the way a claims handler would decide them

Frequently asked questions

Which insurance AI systems do you evaluate?

Customer-facing assistants on the web, in the app, on WhatsApp and on the phone, internal copilots for claims handlers and underwriters, and tools that brokers use to quote or answer clients. We focus on questions where a wrong answer costs money: coverage, exclusions, waiting periods, excesses and claims.

Who decides what the correct answer is?

Your documents first: general and special conditions, product sheets and procedures. Each question cites the clause its answer comes from. Our loss adjusters and claims handlers review the answer key, and your own expert signs it off in a 90-minute session, so the results can't be disputed later.

Do you need access to our systems or policyholder data?

Not to start. The free diagnostic uses only your public assistant and your public documentation, with made-up customer profiles. For internal systems we work under an NDA and a data processing agreement on EU infrastructure, or we give you the questions and you run them yourself.

Does this help with the EU AI Act?

It gives you evidence, not a certificate. AI used for risk assessment and pricing in life and health insurance is high-risk under Annex III of the AI Act, with obligations applying from 2 December 2027. Customer-service assistants are usually not high-risk, but they still have to answer correctly. We provide scores, failures and a run history your compliance team can use. We don't carry out conformity assessments.

How long does it take and what does it cost to start?

The diagnostic is free: we put 40 questions to your public assistant and send you the results within 48 hours. A pilot evaluation, with 150 to 300 questions reviewed by our experts, takes about two weeks and is quoted on scope, at a fixed price agreed before we start.

Find out what your assistant gets wrong

Start with a free diagnostic: we put forty questions to your assistant and, within 48 hours, tell you what fails, why, and what data it needs to improve.