Start a project

Faster claims.
Still a person’s call.

AI agents for insurers, MGAs and brokers: claims intake, policy questions, document extraction and underwriting pre-checks, with a person signing off on every decision that matters.

Experience that carries over.

Our experience includes production conversational AI for a US insurtech: policy questions, claims intake and agent hand-off. The patterns carry over to European insurers, under European rules.

  • Answers grounded in the policy wording.
  • Hand-off to a person, with the full context.
  • Every step logged for audit.
First notice of loss agentIllustrative
  1. Received report and 6 photosMotor claim, reported by web form
  2. Matched policy and coverComprehensive cover, €300 excess
  3. Estimated repair costFrom the photos and repairer rates
  4. Human review requiredEstimate of €18,400 is above the €10,000 auto-approve limit

Illustrative example. Try approving or declining.

The rules you work under.

The regulations that shape an AI system in insurance. A summary for orientation, not legal advice.

Regulations relevant to AI systems in insurance
RegulationWhat it means for an AI system
IDDInsurance Distribution Directive (EU) 2016/97Rules for selling and advising on insurance: act in the customer’s best interests, give clear pre-contract information and establish their demands and needs. An assistant that answers product or policy questions has to stay inside those rules.
Solvency IIDirective 2009/138/ECPrudential rules for insurers, including the system of governance, risk management and outsourcing. Bringing an AI system or an outside engineer into an important function falls under its governance and outsourcing requirements.
DORADigital Operational Resilience Act, Regulation (EU) 2022/2554Applies to insurers since 17 January 2025: ICT risk management, incident reporting, resilience testing and oversight of ICT third-party providers, which includes the AI services you depend on.
GDPRRegulation (EU) 2016/679, Articles 9 and 22Claims often carry health data, a special category under Article 9. Decisions based solely on automated processing with legal or similarly significant effects need safeguards under Article 22, including the right to human intervention.
EU AI Act, Annex III 5(c)Regulation (EU) 2024/1689AI used for risk assessment and pricing of individuals in life and health insurance is high-risk from 2 December 2027. Any agent that talks to customers must already say it is an AI under Article 50.

Agents we’d build for you.

Examples

Illustrative starting points. Each one keeps a person in charge of the decisions that matter.

  • First notice of loss

    Takes the first report of a claim by chat, email or web form, asks the follow-up questions, collects photos and documents, and opens a complete claim file for a handler.

  • Policy questions

    Answers customer and broker questions from the actual policy wording, quotes the clause it relied on, and hands over to a person whenever the answer affects cover.

  • Document extraction

    Reads invoices, repair estimates, medical reports and police reports, extracts the fields your systems need, and flags anything missing or inconsistent.

  • Underwriting pre-checks

    Checks submission data against your appetite and rules and prepares a summary for an underwriter. The underwriter signs off; the agent never binds cover.

  • Complaint triage

    Sorts incoming complaints by topic and urgency, drafts an acknowledgement for review, and routes each case to the right team with its history.

High-risk from 2 December 2027.

Life and health insurance pricing and credit scoring are high-risk under Annex III from 2 December 2027. Customer-facing agents already have a duty: Article 50 has required AI disclosure since 2 August 2026.

Check where your system falls

Common questions.

Can an AI agent make claims decisions?

It can prepare them: gather the facts, check them against the policy wording and draft a recommendation. We build agents so that a person makes the decisions that affect a customer, such as declining a claim or paying above a threshold you set. That also keeps you on the right side of GDPR Article 22.

Is a claims agent high-risk under the EU AI Act?

Annex III lists risk assessment and pricing of individuals in life and health insurance, which is high-risk from 2 December 2027. Claims intake and customer service are not on that list, but an agent that talks to customers must say it is an AI under Article 50, which has applied since 2 August 2026. Our AI Act checker gives a first orientation; it is not legal advice.

Where does policyholder data live?

In your environment. We work in your cloud accounts and EU regions where possible, and personal data is processed under a data processing agreement.

Tell us about one insurance workflow.

Claims, policy questions or underwriting. One workflow is enough to start. We reply within 24 hours on working days.