Picture a three-person independent agency with about 1,400 policies. For them renewal season never really ends. Something renews every working day. This morning's list has eleven: six homeowners, three auto, a small restaurant's package policy, and a contractor's general liability. The homeowners carrier has raised premiums 19% across the board, and at least two of those clients will call before the agency calls them.
A good renewal review takes 20 to 30 minutes per account: open the expiring policy and the renewal offer, compare them line by line, check the claims history, think about what might have changed in the client's life or business, decide whether to remarket, and prepare for the call. Nobody has 30 minutes for eleven accounts every day. So the biggest accounts get the full review and everyone else gets a renewal notice.
This post is a blueprint for changing that ratio: letting AI do the comparison and preparation, so every client gets a prepared conversation and the agent spends their time on judgment.
Why renewals got harder
Two things changed. Premiums went up sharply, and they went up unevenly. The Consumer Federation of America found that US homeowners' premiums rose 24% on average between 2021 and 2024, with far bigger jumps in some states and ZIP codes. Auto followed a similar path as repair costs climbed. When a renewal arrives 20% higher, the client wants an explanation and often an alternative, and a renewal notice in the mail no longer does the job.
The second change is the cost of missing something. Failure to procure appropriate or sufficient coverage remains one of the leading causes of agency E&O claims. Renewal is the natural moment to catch changes: the client added a teenage driver, finished the basement, hired three employees, started taking online orders. If nobody asks, nobody knows, until there's a claim.
In the UK, the FCA's pricing rules have banned "price walking" in home and motor insurance since 2022, so a renewal can't be priced above what an equivalent new customer would pay. That shifts the renewal conversation towards cover and service rather than price alone, which makes preparation matter even more.
The blueprint
Six steps, spread over the three months before each renewal date. The model does the reading, comparing and drafting. The licensed agent does the advising.
- Changes questionnaireAI90 days outA short, line-specific questionnaire goes to the client: new drivers, renovations, new employees, revenue changes, new activities. Answers land in the account file.
- Pull the renewalSystem45 to 60 days outRenewal offer from the carrier download or portal, expiring policy, claims from the last three to five years, notes from last year's call.
- Compare expiring and renewalAIsecondsPremium, limits, deductibles, endorsements added or removed, new exclusions, rating factors that changed. Every difference listed with the page it came from.
- Spot gaps and questionsAIChecks the account against the agency's coverage checklist and the client's answers, and lists questions for the agent. Not recommendations.
- Agent decidesAgent5 to 10 minReads the brief, decides whether to remarket, which options to present, and what to recommend.
- Call and documentAgent30 days outThe renewal conversation. Afterwards the model drafts the file note: what was discussed, offered, accepted and declined. The agent checks and saves it.
Step 1: Ask about changes before you need them
Most agencies ask "anything changed?" on the renewal call itself, when it's too late to remarket or adjust. Asking 90 days out gives time to act on the answers.
The questionnaire should be short and specific to the client's policies. A homeowner with auto gets five questions; a contractor gets questions about revenue, payroll, subcontractors and new types of work. The model can write it from the account's lines of business and remind the client once if there's no reply. A client who answers "we finished the basement and put in a home office" has just told you about two potential coverage issues.
Step 2 and 3: Let the model read both policies
Comparing an expiring policy with its renewal is exactly the kind of tedious, careful reading a language model does well, as long as it shows where each finding came from. The useful output isn't a summary. It's a list of differences.
| What changed | Where it hides | Why it matters |
|---|---|---|
| Premium | Declarations page | The first thing the client will ask about |
| Deductibles, especially wind and hail | Declarations or a separate endorsement | Percentage deductibles can mean thousands more out of pocket after a storm |
| Roof settlement terms | Endorsement changing replacement cost to actual cash value for older roofs | A claim pays far less than the client expects |
| Removed or added endorsements | Forms schedule | Water backup, equipment breakdown or service line cover can quietly disappear |
| New exclusions | Policy forms with new edition dates | A changed form edition is worth reading even if nothing else changed |
| Limits | Declarations | Dwelling limits that haven't kept up with rebuilding costs |
| Rating factors | Carrier's renewal letter or rating worksheet | Explains the increase: territory, claims, credit-based score where allowed, vehicle |
The model should cite the page and form for each line, so the agent can check it in seconds. Anything it can't read clearly should be flagged, not guessed.
Step 4: Questions, not recommendations
Here's where the design matters most. The model compares the account against the agency's own coverage checklist and the client's answers, and lists questions for the agent:
- Teenage driver added and no umbrella policy: discuss liability limits?
- Home in a flood-prone area with no flood policy on file: has flood cover been offered and declined in writing?
- Contractor's revenue up 40%: does the GL exposure basis need updating before audit?
- Restaurant now does delivery: are drivers using their own cars, and is hired and non-owned auto in place?
- Small business taking card payments online and no cyber policy: discuss?
Each question points the agent to a conversation. None of them tells the client what to buy.
Step 5: One page per account
The agent should be able to read the whole brief in two minutes before dialling:
Account: Smith household, homeowners and two autos, client since 2019. Renewal: home +19% (carrier-wide rate increase, no claims), auto +6%. Differences: wind and hail deductible changed from $1,000 flat to 2% of dwelling limit (about $8,400); water backup endorsement unchanged; dwelling limit indexed up 4%. Client answers: daughter got her licence in June, basement finished in spring. Questions for the call: add daughter to auto (not yet on the policy), umbrella discussion, update dwelling limit for the finished basement, explain the new wind deductible. Remarket? Worth quoting home with two other carriers given the size of the increase.
That's the difference between "Your renewal is attached" and a call the client remembers.
Step 6: Write it down
After the call, the model drafts the file note from the agent's quick notes or the call transcript, if the client agreed to recording: what changed, what was offered, what the client chose, and what was declined. The agent reads and saves it. It takes a minute and it's the part most often skipped in a busy week.
What it saves
- Renewals per month
- 120
- Manual prep per renewal, if done properly
- × 25 min
- Hours needed for full prep on every account
- 50 h
- Agent review of a prepared brief
- × 8 min
- Hours with briefs
- 16 h
- Hours freed per month, while every account gets a review
- ≈ 34 h
The real win isn't the hours. It's the accounts that used to get no review at all. Those are where the missed teenage driver and the finished basement live.
Tools that fit
Most independent agencies run an agency management system such as Applied Epic, Vertafore AMS360, HawkSoft or EZLynx, with carrier downloads bringing renewal data in. Comparative raters handle the remarketing quotes. The AI layer sits alongside: it reads the renewal documents and the account, writes the brief and the questionnaire, and saves the output back to the account as an activity or attachment. For carriers that don't send full documents by download, a PDF from the carrier portal works fine.
Client data here includes financial and sometimes health-related information, so the usual rules apply: a provider that doesn't train on your data, processing where you can account for it, and access limited to what the task needs. In the US, the Gramm-Leach-Bliley Act and state insurance data security laws set the baseline. In the UK and EU, GDPR does.
Questions agents ask
Can AI really read carrier renewal documents accurately?
It reads declarations pages and forms schedules well, and it's good at spotting that a form edition changed. It's less reliable on scanned documents with poor quality and on unusual carrier layouts. That's why every difference should come with a page reference. Treat it as a very fast first reader, and check anything that matters.
Should the client get the brief?
No. The brief is internal preparation. The client gets a clear explanation from the agent and, afterwards, a written summary of what was agreed, which the model can draft and the agent approves.
Does this work for commercial lines?
Yes, and the value per account is higher: exposure changes like revenue, payroll, vehicles and new operations matter more, and the policies are longer. Start with small commercial packages, where renewals are frequent and fairly standard.
Will clients find the questionnaire annoying?
Not if it's short and clearly about them. Five relevant questions sent three months ahead, with a line saying why ("so we can check your cover still fits before your renewal"), usually get answered.
Rule of thumb
Every renewal deserves a comparison, a question about what changed, and a note of what was decided. Let the model do the reading and drafting for every account, and spend your own time on the conversations where advice actually changes something.
If renewals are piling up at your agency, tell me which management system you use and roughly how many renewals you handle a month. I'll suggest how a renewal brief could fit your workflow. For a related problem in another profession, see how law firms speed up first contact without giving advice, and how recruitment agencies turn interview notes into usable records.
Building something with AI?
I help small businesses turn ideas into software that pays off. Tell me what you’re working on and get a free first assessment.