AI for insurance agencies: renewals, quotes and complaints

In an agency the AI for Insurance Agencies serves to govern the daily flow: understand whether a message is a request for quote, a variation, a claim or a complaint, prepare the questionnaire on the requests and needs of the customer, keep track of the portfolio deadlines. It's no use choosing the policy over the middleman.

Reason is written in Private Insurance Code. Article 119-ter asks the distributor, prior to the conclusion of the contract, to acquire the information from the contractor to identify requests and needs and to propose only contracts consistent with them; when there is advice, asks for a personalized recommendation with the reasons for the choice. A chatbot that "advises" a guarantee enters this perimeter, and the responsibility remains with the agency.

Here you will find where automation helps an agent or broker, with the rules on complaints and renewals that the system must comply with.

Inform yes, recommend no: where the Insurance Code stops the assistant

The distinction that matters is between information and advice. Paragraph 1 of article 119-ter asks to acquire the information useful to identify requests and needs and to give objective information about the product in an understandable form; paragraph 3 describes the advice as a personalized recommendation that explains why a contract is the most suitable. The first job can be prepared with the help of a system. The second is the work of the intermediary, who responds to it in the first person.

The practical problem is that a linguistic model tends to respond anyway. If a client writes "does kasko or crystal suit me?", the assistant must collect the information and pass the conversation to a person, without choosing. This rule should be written in the instructions and tested on real questions before opening the channel to customers; if the channel is WhatsApp, how does an assistant work on that channel and when convenient is explained in the guide to WhatsApp chatbots for businesses.

Estimates, variations, claims, complaints: sorting agency mail

The day of an agency comes from different channels: email, PEC, WhatsApp, phone, company portals. The first useful automation is a classifier that reads every message, understands what it is and extracts the data that is needed. It is the same principle as the ticket management with AI, applied to a policy portfolio.

The right column weighs more than the others. The system may also prepare the answer to a frequent question, but a cancellation or complaint must arrive at the right desk immediately. For the answer part there is the guide on how AI responds to business emails, and the demo of a AI Agent for Business Emails shows how to set the sorting.

RC car renewals: no tacit renewal, and the expiration notice is up to the company

For the RC car the Article 170-bis of the Code It sets precise rules. The contract lasts a year, or a year more than a fraction if the policyholder so requests, is terminated at the expiry and cannot be tacitly renewed; the termination at the expiry also applies to ancillary risks guaranteed in the same contract or in one stipulated at the same time. The company must notify the policyholder of the expiration with at least 30 days notice and keep the warranty up to the effect of the new policy, no later than the fifteenth day after the expiration.

For the agency each deadline is therefore a contact not to be missed. A system connected to the management system can prepare every week the list of expiring policies, report customers who have not responded to the first message and write a draft with the data to verify: vehicle, drivers, any changes. The message starts after a person has checked it, and does not promise conditions that the company has not yet confirmed.

The same applies to unanswered quotes. The reminder sequence described in the guide to customer follow-up automation It works in the agency if the rules on when to stop, and on who to call back, the owner decides.

Complaints: recognize them immediately and send them where the procedure wants them

On complaints the rules change with the section of the register of intermediaries. As the IVASS explains in FAQ on dealing with complaints from intermediaries, for agents (Section A) and direct producers (Section C) complaints are handled by the leading companies, even when they concern employees and collaborators. Brokers (section B) have their own procedure and record complaints about the models under ISVAP regulation 24/2008.

On the customer's side, the first step is to write to the company's complaints office, which must respond within 45 days; without answer, or with an unsatisfactory answer, you can go to IVASS. From 15 January 2026 you can also resort to the Insurance Referee: the appeal, against a company or an intermediary, only occurs online after a complaint has been unanswered within 45 days or with an unsatisfactory response.

Life and health policies: the data that a chatbot must not collect

In life, injury and illness policies often enter health information. The Article 9 of the GDPR prohibits the processing of health data except in specific cases, such as explicit consent: an assistant who asks for pathologies or therapies in chat creates a treatment that the agency must know how to justify and protect. The prudent way is to leave health questionnaires and medical documents to the channels provided by the company.

Then there is the level of the company. The Annex III to the AI Act It ranks high-risk systems used to assess risks and determine prices for individuals in life and health insurance; with the Regulation (EU) 2026/1744 the obligations for these systems shall apply from 2 December 2027. For an agency the practical consequence is simple: pricing and risk taking remain in the company's tools, and a home-built score on those customers is a bad idea. A picture of the categories is in the guide to risk levels of the AI Act; for the privacy part there is the article on GDPR and artificial intelligence in the company.

Two examples: an agency with two mandates and a broker for SMEs

Monday morning of a multi-mandatory agency

Think of a provincial agency with two warrants and three collaborators. On Mondays the box collects weekend requests: in this example about forty messages between car RC quotes, photos of a pad, a cancellation via PEC and an email about a refund that does not arrive. The system sorts them by type, opens the practices and recognizes in the refund email a possible complaint: it reports it to the holder, who forwards it to the company's complaints office according to the mandate procedure.

For each quote the questionnaire on the needs starts with the customer, and the collaborator finds the card already filled out when he calls back. No response came out of the agency without a person reading it.

The renewal of a multi-risk for a small company

A broker follows a manufacturing SME who renews the multi-risk policy. Three offers from different companies arrive, each with its own lexicon for franchises, caps and exclusions. The system puts them in a table with references to the conditions of each and reports the differences with respect to the current policy.

The recommendation is written by the broker. If he presents his advice as based on an impartial analysis, paragraph 4 of article 119-ter asks him to base it on a sufficient number of contracts available on the market: the comparison prepared by the system documents a part of that work, the choice remains his.

The Most Expensive Mistakes for an AI-Used Intermediary

• Letting the chatbot answer "yes, you're covered" after reading the general conditions: whether a claim will be paid depends on the facts and the valuation of the liquidator.
• Have the system choose between two guarantees when the client asks for advice.
• Treating a complaint as a normal request for assistance, or leave it in the queue for days.
• Collecting in chat health data for life or disease policies.
• Loading the portfolio list into tools that don't offer written agreement on how they treat the data, perhaps to do a renewal analysis first.

From which process to start in the agency

The safest starting point is internal: mail sorting and renewals schedule. Neither of them talk to customers without passing by a person, and both are easily measured: messages taken care of in the day, expiring policies contacted before the date, complaints recorded the same day. The assistant who talks to customers comes later, with a written perimeter: gathers information, does not advise.

Many of these flows are common to other intermediaries and firms working for companies: you find them collected in the overview dedicated to B2B Services.

If you want an agent who sorts the agency box, recognizes complaints and claims, and prepares ballots for collaborators without ever answering for you on the covers, Let's start with how your messages arrive today.