AI agent for insurance agencies with renewals tracked to the day
The book of policies sits in a few insurer portals and in the agency management system, and renewal dates are watched by a spreadsheet and somebody's memory. We pull that into one place: a renewal calendar that reminds you on its own, document packs checked against a list of gaps, and an AI agent answering from your own wordings, with the section named. The decision stays with a person.
- 14 daysfrom the end of the workshop to the first automation running, built on the data you already receive
- EUR 3,500net per week of workshop, a published price we do not negotiate. The rollout is quoted after it
- 5 to 10 secondshow long the agent takes to find an answer in your own policy wordings and name the document and section it came from
The problem
Why does a renewal fall out of the calendar?
One client has motor cover with one insurer, property with a second and the company fleet with a third. Each of them has its own portal, its own export format and its own way of showing the expiry date. The book of business only exists in one piece when somebody stitches it together by hand, and between one stitching and the next the renewals are watched by a spreadsheet and by whoever happens to remember.
The second thing is documents. A complete pack for a policy takes three rounds of email, because the gap only becomes visible when somebody sits down to issue the cover. A claim works the same way, only it costs more: photos arrive with no report, the report arrives with no claim number, and looking after the client after a loss comes down to explaining what stage the file is at. Commissions get reconciled by hand, after the month is closed.
The third layer is formal and there is no way around it. You work with personal data and with claims data, and Article 22 of the GDPR gives a person the right not to be subject to a decision producing legal effects that is taken solely by automated means. That is why the agent does not assess risk, does not issue a policy and does not settle a claim. This is not legal advice, and we set out the legal basis separately in the guide to AI and the GDPR.
- A renewal reminds you it exists when the client rings to ask about the new premium
- A complete document pack for a policy collected over three rounds of email
- Service after a loss is mostly explaining what stage the claim is at
- Commissions reconciled by hand from several files, only after the month closes
Why now
What changed, so that this pays off today
The agency management system, the mailbox, the calendar and the spreadsheets now have interfaces or at least predictable exports, so pulling them into one view is counted in working days rather than in a quarter. Where an insurer gives you no interface, we work with what reaches you anyway: the statement file and the email.
The second change matters more. A rule will carry a date from one system to another, but it will not answer the question of whether a given loss falls inside the cover written a year ago. A language model searches your own policy wordings, endorsements and procedures and answers in a sentence, naming the document and the section the answer came from. Not from the internet, from your files.
The third thing is a boundary we set ourselves and state out loud. The agent prepares the material and points to the source, while risk assessment, issuing a policy, settling a claim and choosing a product for the client stay with a person. The same person is accountable as before, the one who puts their name to it.
What changes
What is visible after the first month?
No renewal falls out of the calendar
Expiry dates from insurer portals and from the agency management system land in one calendar. The reminder goes to the account handler and separately to the client, on two thresholds, with nothing keyed in by hand.
Two reminder thresholds, 60 and 30 days before cover ends
A complete document pack at the first attempt
Every class of policy and every type of claim has a list of required documents. The system compares it with what actually arrived and sends the client the list of gaps straight away, not a week later.
The list of gaps sent at first contact instead of a third round of email
One view of the client's book
Policies, expiry dates, premiums and claims history in one place instead of five browser tabs. The conversation with the client starts from the facts rather than from logging in.
1 view instead of 5 portals open side by side
Claim status reaches the client on its own
A change of stage goes out to the client the moment it happens. A phone call asking what happens next stops being the main reason a client rings the office.
A notification on every change of stage, with no client having to ask
An answer from your wordings, with the section named
The AI agent searches the policy wordings, the endorsements and your own procedures and answers in a plain sentence. Every answer names the document and the section, so it can be checked before you speak to the client.
An answer in 5 to 10 seconds, with the document and section named
Commissions reconciled before anyone asks
Statements from insurers are matched against your policy register and the differences are written out. Instead of building a table from scratch, you review the list of entries that do not agree.
A reconciliation run every night, with a list of items to query
Proof
Do you use this yourselves?
We do not have a brokerage that would agree to publish its own numbers, and we are not going to get around that with an anonymous example. We show the only thing we can show in full: an AI agent running in production at our founders' company, built with the same method. It is a services business billed client by client, with deadlines nobody is allowed to miss. Below is six months of measurement, together with what those numbers do NOT prove.
Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof.
JustBottheir agentJustIdea Agency, a performance marketing agency of around 30 people
- 23
- people from the company work with the agent
- 38
- systems wired in permanently
- 112
- processes run without a human
- 475
- thousand lines of the agent’s code
An agent wired into the CRM, the task system, the mailbox, the data warehouse and the ad platforms. First commit 1 March 2026, today 23 people use it.
8 of 38 systems it works in
Google Ads
Meta Ads- HubSpot
- ActiveCollab
- Ahrefs
GA4
Search Console
BigQuery
- Before
- In the twelve months before launch the agency ran at a margin of minus 4.42 percent. Across a twenty-month window there was not a single run of three profitable months.
- After
- May, June and July 2026 are three profitable months in a row, the first such run in the whole window. Revenue barely moved, so the change sits on the cost of delivery, not on sales.
Margin up 5.55 percentage points year on year, with revenue up 1.5 percent
And in your agency
What if this does not work for us?
Four sentences we hear most often from agencies and brokers, and what we say back to them. If yours is a different one, say it on the consultation.
Every insurer has its own portal and none of them will give us an interface.
We are not promising you an interface where none exists, and we do not build anything that logs into somebody else's portal on your behalf. The starting point is what reaches you anyway: the agency management system, the statement file, the export, the confirmation email. That is what the renewal calendar and the book view are built from. In the workshop we say plainly which data cannot be collected this way.
We will not agree to AI issuing a policy or settling a claim.
Neither will we, and that is how we build it. The automation takes over what a person does with a keyboard today: watching the date, completing the document pack, sending the status, reconciling commissions. Risk assessment, issuing the policy, the decision on a claim and choosing a product for the client stay with a person. That also follows from Article 22 of the GDPR, which gives the client the right not to have a decision with legal effects taken solely by automated means.
Our book sits in a spreadsheet, not in any system.
The spreadsheet is a starting point, not an obstacle, and in smaller agencies it is what we meet most often. We begin by tidying the columns and the dates, then add what comes out of the agency management system. Replacing a system is a separate decision and a separate project, while the first automation is meant to exist 14 days after the workshop ends.
This is clients' personal data and claims data.
We agree the scope in the workshop and write it down: the model sees named sources and named fields needed for the task, not the whole agency database. You hold a data processing agreement with the model provider, access is set per person, and queries stay in a log. If you prefer it, part of the flow runs on your own infrastructure.
- CRM and sales
- HubSpot, Pipedrive, Salesforce
- Sources of policy data
- the agency management system, insurer portals, statement files, exports
- Mail, calendar and documents
- Google Workspace, Microsoft 365, a shared inbox, drives, scans
- Billing and commissions
- Xero, QuickBooks, Fakturownia, commission spreadsheets
- Data and reporting
- Google Sheets, BigQuery, SQL databases
- The automation layer
- Make, n8n, custom API integrations
What exactly we automate in an insurance agency
A typical split of the work in an agency and in a brokerage, before a rollout and after it.
| Area | Today | After the rollout |
|---|---|---|
| Renewal calendar | a spreadsheet and the handler's memory | thresholds at 60 and 30 days, a reminder to the handler and to the client |
| Documents for a policy | gaps show up at the point of issue | the list of gaps sent back at first contact |
| Notifying a claim | documents arrive in batches | a checklist per type of claim, gaps written out at once |
| Claim status | the client rings and asks | a notification on every change of stage |
| The client's book | several portals and a spreadsheet next to them | one view: policies, dates, premiums, claims history |
| A question about the scope of cover | somebody searches the wording file | an answer in 5 to 10 seconds, with the document and section |
| Commissions | reconciled by hand after the month | a reconciliation every night, with a list of items to query |
The scope of the first rollout is agreed in the workshop. We take the three items with the highest return, not all of them at once.
How we work
Five steps from a renewal spreadsheet to an agent
Process first, tools second. Every step ends with something you can judge before you decide on the next one.
Process workshop
One week. We walk the path of a policy from the client enquiry through issue and renewal to the claim and the commission statement. We come out with a map and a list of the places where the same date or the same document is created twice.
Boundaries in writing
Before we build anything, we write down what the automation does not do: it does not assess risk, does not issue a policy, does not settle a claim and does not advise the client on which product to buy. We also settle what data the model sees and who has access to what.
The first automation live
Fourteen days to something that works. In an agency we usually start with the renewal calendar, because it shows an effect in the first week and needs no insurer's permission.
Agent pilot on your documents
The agent that answers from the wordings and the procedures goes to a pilot first, on real questions from the team. There is one condition for going to production: when the documents give no basis for an answer, the agent says it does not know rather than guessing.
Measurement and the next areas
At 30 and at 90 days we check the indicators agreed before the start. Only then do we decide whether to move into the next area, claims or commissions for example.
Before and after
How the week changes for an insurance account handler
Before
- Monday goes on checking portals to see whose cover is ending
- The client raises the renewal himself, usually a week before expiry
- Documents for a policy collected over several rounds of email and phone calls
- After a loss, a call from the client with one question: what stage is this at
After
- Monday starts with a finished list of renewals for the next 30 days
- The reminder went to the client and to the handler before anyone thought about it
- Gaps in the documents written out at first contact, in a single message
- The client knows the stage, because the notification went out the moment it changed
What does not change is who assesses the risk and who decides on the policy. What changes is how much time goes on watching dates and chasing paperwork before anybody starts talking to the client.
For the owner of an agency or a brokerage
What the person signing this off gets out of it
The first thing shows after a month: the book stops being one person's knowledge. You can see how many policies expire in each month, which of them already had a reminder sent and where the documents stalled. That is the basis for a conversation about headcount and about how many more clients you can take on.
The second is quieter, and it matters most when somebody leaves. Dates, contact history and documents stay with the agency rather than in somebody's mailbox and somebody's spreadsheet. Access is written down per person and queries to the agent stay in a log, so it is clear who saw what.
- A renewal list for the next 30, 60 and 90 days that builds itself
- Documents and contact history held by the agency, not in a private inbox
- Access to client data written down per person, with a log of queries
- The boundaries of the automation written down before the rollout, not after the first complaint
What it costs
The workshop has a published price, the rollout is quoted after it
The process workshop costs EUR 3,500 net per week and runs from 1 to 7 weeks, and we give that price before the first conversation, because an agency has a right to know what it is agreeing to. For a single agency one week is usually enough; for a brokerage placing business with several insurers it is sometimes two. We quote the rollout after the workshop, once it is clear how many data sources are in scope and what state they are in.
The cost of using the model and the tools you pay straight to the providers, with no margin from us, and a single query costs the provider roughly EUR 0.005 to 0.07. With client data that matters formally as well: the contract is between you and the provider, so it is clear who entrusts what to whom. There is more about the week itself on the process workshop page, and the running cost of scenarios in n8n or Make works the same way.
Process workshop
EUR 3,500 net
per week, and in an agency usually one to two weeks
A self-contained piece of work. It ends with a map of the policy journey and a 30 day plan you can carry out without us.
- A week of work inside the agency and inside the data sources
- A map of the policy journey from enquiry to renewal and claim
- Written boundaries for the automation and the scope of access to client data
- The three automations with the highest return, costed before the start
- Most popular
Workshop and first rollout
Quoted after the workshop
scope follows from the number of data sources and the state they are in
For an agency that wants a renewal calendar running 14 days after the workshop ends.
- Everything in the workshop
- The first automation live in 14 days
- A renewal calendar wired into the agency management system and the mailbox
- Training for the person who runs this on your side
- Measurement at 30 and at 90 days
AI agent for the agency
Quoted after the workshop
for teams where questions about the scope of cover come up every day
BOND, the company AI agent, answering from your wordings and procedures, always naming the document.
- Everything in the rollout
- Answers from wordings, endorsements and procedures, with the section named
- A summary of the client's book: policies, dates, premiums, claims
- Access per person, with a log of queries
The workshop ends with a document that stands up without us: a map of the process and a 30 day plan you could carry out with your own people. If after the workshop we decide there is nothing worth automating in your agency, we say so plainly and do not propose a rollout.
Frequently asked questions
What does automation mean in an insurance agency?
It means a system taking over the work an account handler or the back office does today: watching the dates cover expires, completing document packs for policies and for claims, sending the client the status of a file and reconciling commissions against insurer statements. You usually hold the data already, it simply sits in several places at once. Automation is about bringing it into one flow, not about buying another system.
Will the AI agent issue a policy or settle a claim?
No, and that is not what we build it for. The agent does not assess risk, does not issue a policy, does not settle a claim and does not advise the client on which product to buy. It prepares the material, points to the section in the wording and shows what is missing from the documents. A person makes the decision and puts their name to it, exactly as before the rollout.
Where does the agent get an answer about the scope of cover?
From the documents you point it at: policy wordings, endorsements, your own procedures and notes. Every answer carries the name of the document and the section, so checking it takes under a minute. If the documents give no basis for an answer, it has to say it does not know. We test that condition in the pilot, and until it holds we do not go to production.
How does the renewal calendar work?
We collect expiry dates from the agency management system, from statement files and from the spreadsheet you already keep, and hold them in one place. The reminder goes out on two thresholds, usually 60 and 30 days before expiry, separately to the handler and to the client. We do not promise you a higher retention rate, because that is decided by the conversation and the premium. We promise that no date gets lost in a spreadsheet.
Will this renew more policies for us?
We will not give you a percentage, because we have no measurement from the insurance market and we are not going to invent one. What we take responsibility for is narrower and checkable: no renewal falls out of the calendar, the reminder goes out on time, and the client gets an answer faster because nobody is hunting through five portals. The rest depends on the conversation with the client.
What about clients' personal data and claims data?
We write the scope down in the workshop: the model sees named sources and named fields, not the whole agency database. You hold a data processing agreement with the model provider, access is set per person, and queries stay in a log. Decisions about a client stay with a person, which is also the answer to Article 22 of the GDPR. The details are in our guide to AI and the GDPR, and this is not legal advice.
Do we have to change the agency management system?
No. We read data out of it and leave it where it is. Replacing a system like that is a project measured in quarters and a separate decision, while the first working automation is meant to exist 14 days after the workshop. If the system exposes no interface, we say so in the workshop and show what can be done with the export, the mailbox and the spreadsheet.
How long does it take and what does it cost?
A week of workshop at EUR 3,500 net, then 14 days to the first working automation. A full agent answering from your wordings and procedures usually takes around 2 months, because it needs a pilot on real questions. The rollout is quoted after the workshop, once we know the number of data sources and their state. We measure the effect at 30 and at 90 days.
Let us start with a single renewal
The free consultation takes 30 minutes. We walk the path of one of your policies from the client enquiry to the renewal and tell you whether we see anything there worth automating. If we do not see it, we say so on the call. You write directly to Pawel, with no reception desk and no ticket in a system.
We run at most two rollouts at a time, so that someone is genuinely sitting on each of them.
Book a free consultationPS. If a renewal this month announced itself only through a phone call from the client, that is exactly this problem. We will not issue the policy for you and we will not settle the claim, but no date will disappear from the calendar and the client will get an answer faster.