In short
- Good candidates are document-heavy steps: structuring submissions, comparing quotes, intake and sorting of claim notifications, checking policy documents against the order, preparing renewals and file notes.
- Cover decisions, claims acceptance, underwriting, pricing and advice stay with people. The AI prepares, a qualified person decides.
- AI systems for risk assessment and pricing of natural persons in life and health insurance are high-risk under Annex III of the AI Act. We do not build them.
- Customers ask AI assistants which broker or insurer to trust. Firms with register entries, clear pages per line of cover and consistent profiles are easier to name.
Typical workflows in insurance
These are typical processes, not client projects. In each, the AI prepares, a qualified person checks, and the result goes into the broker or policy system with a log entry.
1. Submissions and quote comparison
A broker collects the client’s details from emails, questionnaires and schedules, sends them to several insurers and receives quotes as PDFs in different layouts.
- Input: client emails, questionnaires, asset and sums-insured schedules, quote documents from insurers.
- AI prepares: structures the details into the submission format each insurer expects, flags missing information and lays the returned quotes side by side: cover, limits, excesses, exclusions and conditions with the page each comes from.
- Adviser checks: the submission before it goes out, and the comparison against the client’s needs.
- Output: a submission pack and a comparison table as drafts. The recommendation is the adviser’s.
2. Claim notifications
Notifications arrive by email, telephone note or portal, with photos, invoices and police or expert reports attached.
- Input: the notification and its attachments.
- AI prepares: reads policy number, date and type of loss, matches the policy, checks cover period and insured parties against the policy record, lists missing documents and drafts the acknowledgement and the request for the missing items.
- Claims handler checks: every cover question. Whether the claim is accepted, reduced or declined is decided by a person.
- Output: a claim file opened with the documents sorted, plus the draft letters.
3. Policy documents against the order
New policies, schedules and endorsements are read to see whether they match what was ordered.
- Input: the order or proposal and the document received from the insurer.
- AI prepares: compares names, addresses, sums insured, periods, excesses and special conditions and lists every deviation with both passages.
- Adviser checks: each deviation and asks the insurer to correct it.
- Output: a short check report per document and a draft query to the insurer.
4. Meetings, file notes and renewals
Advice has to be documented, and renewals need the history of the policy gathered from several places.
- Input: notes or a recording of the client meeting, the policy file and the claims history.
- AI prepares: a draft file note that records the client’s demands and needs and the options discussed, and a renewal pack with changes and claims history.
- Adviser checks: the content and its accuracy, because the advice and its record are the firm’s responsibility.
- Output: the file note and the renewal pack as drafts, saved in the broker system.
What stays with people, and what we do not do
- Cover, claims and prices: accepting or declining a claim, underwriting a risk and setting a price are decisions by qualified people.
- Advice: recommending a product is the adviser’s regulated activity. The AI supplies facts and drafts.
- Risk assessment and pricing of individuals in life and health insurance: these systems are high-risk under Annex III of the AI Act. We do not build them, see risk categories.
- Fraud accusations: the AI may flag documents that look inconsistent. It never decides that someone has acted dishonestly.
- Customer chat without hand-over: a chatbot answers questions about documents and processes, says clearly that it is an AI and hands over to a person for advice, claims decisions and complaints.
Regulation and data protection
The table lists the main supervisory sources as an orientation. Which rules apply to a firm depends on its status, licence and markets.
| Jurisdiction | Basis | What applies in essence |
|---|---|---|
| UK | FCA, our approach to AI and Consumer Duty | The FCA does not plan AI‑specific rules. Existing frameworks apply, including the Consumer Duty and the accountability of senior managers. Customer communications must stay clear and not misleading, whichever tool prepared them |
| Ireland | Central Bank of Ireland, Consumer Protection Code | Regulated firms must act in the best interests of consumers. Responsibility for outcomes stays with the firm when processes are automated |
| EU | Insurance Distribution Directive (EU) 2016/97 and the EIOPA Opinion on AI governance and risk management | Distributors must act honestly, fairly and professionally in the customer’s best interests and assess demands and needs. EIOPA expects insurers and intermediaries to govern and document AI use in proportion to its risk |
| Germany | § 6 VVG and § 61 VVG | Insurers and intermediaries must advise the customer, document the reasons for their advice and hand over the documentation |
On top of this, Art. 22 GDPR limits decisions based solely on automated processing that significantly affect a person, and Art. 28 requires a data processing agreement with every provider. Art. 4 of the AI Act requires sufficient AI literacy among staff, and Art. 50 requires AI chatbots to identify themselves. This overview is general information, not legal advice. Check the current texts with your regulator and your compliance officer.
Systems you probably use
UK brokers typically work with broker systems such as Acturis, Applied Epic or Open GI, and carriers with platforms such as Guidewire or Duck Creek. In Germany, data exchange between insurers and brokers follows the BiPRO standards. We connect through the interfaces these systems offer (API, export, email or a watched folder) and work alongside them where none exists. The names are for orientation only and do not imply any partnership.
How AI assistants recommend insurance firms
Customers ask ChatGPT, Gemini or Google’s AI Overviews for “a broker for commercial property cover in Manchester” or “who handles landlord insurance”. The answers draw on the firm’s own pages, register entries, trade body listings, the Google Business Profile and reviews. Because insurance is regulated, assistants and users alike look for firms that are visibly authorised.
- One page per line of cover: who it is for, what is typically covered, what is excluded and how claims work.
- Register entries that match: firm name, address and permissions identical to the regulator’s register, for example the FCA Register in the UK or the Vermittlerregister in Germany.
- The same details everywhere: website, Google Business Profile, trade body listings such as BIBA or Brokers Ireland, and directories.
- Clear, fair content: no exaggerated claims, in line with the rules on customer communications.
Is it worth automating? Eight questions
- Do you re-key client details into several insurer portals or forms?
- Do you compare quotes in a spreadsheet that someone fills by hand?
- Do claim notifications arrive in many formats and take time to sort?
- Are policy documents checked against the order only when someone has time?
- Are file notes written after the meeting from memory?
- Do renewals start late because the history has to be gathered first?
- Is there a written rule on what may go into AI tools and who checks the result?
- Can you name one task where a draft that needs checking would already save time?
Three or more yes answers usually point to a workflow worth testing. Question 7 is the prerequisite.
Sources
- UK: FCA, AI and the FCA: our approach and Consumer Duty
- Ireland: Central Bank of Ireland, Consumer Protection Code
- EU: Directive (EU) 2016/97 (IDD): EUR-Lex; EIOPA, Opinion on Artificial Intelligence governance and risk management
- Germany: § 6 VVG, § 61 VVG
- Regulation (EU) 2016/679 (GDPR), Art. 22 and 28: EUR-Lex; Regulation (EU) 2024/1689 (AI Act), Art. 4, 50 and Annex III: EUR-Lex