Automate the renewal timeline, not the advice. Trigger tasks at 90, 60, 30 and 14 days before expiry, pre-fill renewal questionnaires from last year's data, let AI draft reminders and renewal reports, and score each client's risk of leaving so brokers phone the ones most likely to go. Clients stay because conversations start earlier and go better, not because they get more emails.
All of this depends on the data being right: expiry dates, contacts, and what has changed in each client's business. An AI-drafted email asking a video production company about "your fleet" after they sold the vans in spring does more damage than silence, because it tells the client nobody has looked at their file. Clean the policy data before you automate anything, and let the questionnaire keep it clean from then on.
The renewal calendar, stage by stage
Write the timeline down as triggers your broker system or automation can act on. A filled-in version for commercial lines, to adjust for your classes and any notice periods your regulator or insurer agreements require:
| Days to expiry | What happens | Automated | Human |
|---|---|---|---|
| 120 | Large or complex accounts: strategy review | Task created for account handler | Broker decides whether to market the risk |
| 90 | Pre-filled questionnaire sent | Email and form, drafted from last year's data | Handler glances at the draft for obvious changes |
| 75 | Questionnaire not returned | Friendly chaser | None |
| 60 | Still not returned; or returned with changes | Task: phone the client | Handler phones; reviews changes |
| 45 | Terms requested or received | Comparison tasks created | Broker reviews terms and wording changes |
| 30 | Renewal report sent | Report drafted from terms and comparison | Broker edits, approves and calls high-risk clients |
| 14 | No decision yet | Reminder | Handler phones |
| 7 | Final reminder | Reminder stating what happens at expiry | Broker phones any high-value client |
| 0 to +7 | Bound, or lapsed | Confirmation, or a lapse follow-up task | Broker calls lapsed clients personally |
The pattern to notice: automation does the reminding and the drafting, and every point where a client might be lost has a named person making a phone call.
Clients with several policies break a calendar keyed to single policies. Picture an illustrative architecture practice with professional indemnity renewing in February, office contents in June and cyber in October. A per-policy trigger sends it three questionnaires a year, each asking for staff numbers, fee income and the same premises details. By October the practice manager has stopped replying, and the cyber renewal scores high on lapse risk for a reason the brokerage created. Key the questionnaire to the client, not the policy: one "confirm or change" form at the earliest expiry, with a short section per policy, and later renewals pre-filled from that year's answers. Where it makes sense, the call at day 60 is also the moment to suggest aligning expiry dates.
Pre-filled questionnaires clients actually return
A blank renewal questionnaire is a chore, and chores get ignored. A "confirm or change" form built from last year's answers takes a client five minutes. AI is useful here in turning last year's proposal data into short, specific questions:
From last year's proposal data below, write a renewal questionnaire in
"confirm or change" format. For each item, show last year's answer and
ask the client to confirm it or give the new figure. Group into:
the business, people, premises, turnover, equipment and vehicles,
claims and incidents, anything new. Plain English, no insurance jargon.
Add at the end: "Has anything else changed in how you work that we
should know about?" Do not add questions not supported by the data.
Proposal data: [paste]An illustrative extract for a video production company, and the handler's fixes:
Vehicles: Last year you had 2 vans (registered to the business).
Is this still correct? [Confirm] [Change: ____ ]
Equipment: Last year's total equipment value was 180,000.
Is this still correct? [Confirm] [Change: ____ ]
People: Last year you had 12 employees and used around 20 freelancers.
Is this still correct? [Confirm] [Change: ____ ]The handler's diary note from a mid-term call says the vans were sold in April and the company now hires vehicles per shoot. That goes into the form before it is sent, as "We understand you sold your vans in April and now hire vehicles. Is that right?", which also opens a question about hired-in vehicle cover. The equipment line gets a nudge ("Many production companies add kit during the year; please include anything bought since last renewal"). The AI can't know what the handler knows; the 30-second glance before sending is where that knowledge gets in.
When the form comes back with gaps, the missing-information request is another short draft. For the same client, with the equipment line left blank and a note saying "list attached", the AI's first draft asked for the equipment value and a list of items bought during the year. Both were in the attached spreadsheet. Include attachments in what the AI reads before it drafts, and tell it to ask only for what is missing from the form and its attachments together. Clients forgive one follow-up question; being asked for something they've already sent tells them nobody opened it.
A lapse-risk score that tells you who to phone
Brokers can't phone every renewal, and they don't need to. A simple, visible score shows which clients deserve a call before the renewal report rather than after. Keep the rules transparent, so anyone can see why a client scored high:
| Signal | Points |
|---|---|
| Premium rise over 10% | 2 |
| Premium rise over 20% | 3 more |
| Claim during the year, especially if declined or disputed | 2 (declined: 4) |
| Material cover reduced at renewal | 3 |
| No contact with the client in 12 months apart from renewal | 2 |
| Only one policy with you | 1 |
| Complaint or service issue during the year | 3 |
| Questionnaire not returned by day 60 | 2 |
| Key contact changed at the client | 2 |
Six points or more means a broker phones before the renewal report goes out. Three filled-in illustrations:
- A translation agency: premium up 14% (2), single policy (1), questionnaire returned. Score 3: renewal report as normal, with the reason for the rise explained clearly.
- An IT support firm: premium up 22% (5), ransomware sub-limit halved (3), no contact all year (2). Score 10: a broker calls at day 45 with alternatives in hand.
- A PR consultancy: premium flat, new managing director since spring (2), questionnaire not returned (2), claim settled smoothly (2). Score 6: a call to introduce yourself to the new MD, which is also a chance to review cover.
AI can help assemble the score by reading the year's emails and notes for complaints, contact changes and claim frustration, but have it quote the line it relied on, so the handler can check. The rules themselves should be yours, and simple enough to explain to a colleague in a sentence each.
The quote requirement is what catches misreadings. An illustrative extract from the AI's scan of one client's year:
Complaint or service issue: YES (3 points)
Evidence: "we've now waited three weeks for the certificate"
(email to handler, 14 February)
Complaint or service issue: YES (3 points)
Evidence: "honestly this has been a nightmare from start to finish"
(email to handler, 2 July)
Key contact changed: NO evidence foundThe first is a real service failure. The second, read in full, was the client describing their own supplier's late delivery, with no complaint about the brokerage at all. Before quotes were required, a scan like this pushed a satisfied client over six points, and a broker opened the renewal call with an apology the client didn't understand. With the quote in front of them, the handler deletes the second line in ten seconds.
Renewal reports written in minutes, checked properly
The renewal report is where the broker earns the renewal. AI can draft it from the terms, the premium comparison and the wording review, so the broker spends their time on the recommendation rather than the formatting. If the insurer has changed the wording, run the comparison first using the two-pass wording comparison, and feed the corrected results into the report prompt, never the raw terms alone.
A good renewal report opens with the answer: renewal premium, the change from last year and why, any material change in cover, and the broker's recommendation. Then the detail. Ask the AI for that order explicitly; left alone, it writes chronologically and buries the premium on page two.
Messages AI should draft, and ones it shouldn't
| Message | AI's role |
|---|---|
| Questionnaire, chasers, reminders | Drafts; sends automatically once proven |
| Missing-information requests | Drafts from the gaps in the form; handler approves |
| Premium increase explanation | Drafts from the broker's notes; broker approves |
| Renewal report | Drafts; broker edits and signs off |
| Insurer declining to renew | None; broker phones first, then writes |
| Renewal after a disputed claim | None; broker handles personally |
| Complaint or cancellation | None |
The premium increase explanation is the draft most worth checking line by line, because the AI will supply a reason if you don't give it one. The broker's notes for an illustrative bakery with two shops: "Insurer applying about 8% across this class. Escape of water claim in March, settled. Sums insured up after questionnaire: stock and equipment raised to cover the new oven." The AI's draft said the premium had risen "in line with increases seen across the insurance market this year, driven by inflation and rising claims costs", before mentioning any of the three actual reasons. That opening is a generalisation nobody at the brokerage has checked, and it hides the one reason the client controls. The approved version leads with the client's own changes: "Most of the increase comes from the higher sums insured you gave us for the new oven and extra stock, so you're properly covered. The insurer has also applied a rate rise of about 8% across this type of business, and the March claim is taken into account." Tell the prompt to use only the reasons in the notes, in order of size.
A before and after for the seven-day reminder. Before: "Your policy is due for renewal on 1 October. Please contact us to renew." After, drafted from the client's record and approved once as a template:
Your professional indemnity and cyber cover with us ends on 1 October. We sent renewal terms on 1 September: premium up 6%, with no reduction in cover. If you'd like to go ahead, reply "renew" and we'll confirm the same day. If you have questions, your broker is free on Thursday or Friday for a quick call. If we don't hear from you, you won't be insured from 1 October.
It says what, when, how much and what happens next, and makes renewing a one-word reply. That last line isn't pressure; it is information the client needs.
Connecting the broker system, a spreadsheet and an automation tool
Start with what your broker management system already does. Most have diaries or workflow tasks keyed to expiry dates; use them for the calendar before buying anything. Add an automation platform for the gaps, usually the AI drafting and anything that crosses systems.
- A weekly expiry export from the broker system to a spreadsheet, listing every policy expiring in the next 120 days with its contact and last year's key data.
- A Make scenario or Zap that reads the sheet, creates the right stage task, and drafts emails into the handler's inbox. Zapier Professional includes 750 tasks for $19.99 monthly on an annual plan, a task being each successful action step; Make has a free tier of 1,000 credits a month, with paid tiers from about $9.
- A CRM, if you use one for client communications. HubSpot users should know that the flexible workflow builder starts at the Professional tier; Starter's automation is limited to simpler actions around forms and emails.
- A business AI plan for drafting (ChatGPT Business or Claude Team, both of which exclude business content from model training by default), or the AI step inside your automation platform.
Most errors in renewal automation come from the data, not the AI: duplicate clients, stale contacts, expiry dates keyed wrongly. Cleaning customer records before adding AI is worth doing first, and the same deadline-tracking logic appears in tracking contract renewals and deadlines with AI.
A commercial brokerage with 900 renewals a year
An illustrative brokerage has four account handlers and two brokers, with about 900 commercial renewals a year, roughly 75 a month. Before automating, handlers keep their own spreadsheets; questionnaires go out anywhere from 30 to 90 days before expiry; renewal reports are written from scratch; brokers hear about unhappy clients when they don't renew.
After three months of set-up: every renewal gets a pre-filled questionnaire at day 90; about a fifth score six or more on the lapse-risk rubric, which means a manageable 15 or so broker calls a month; renewal reports take around 20 minutes to review instead of an hour to write. Those times are estimates to test against your own, not promises.
The data entry that feeds the process matters too. If much of the handlers' week goes on re-typing client data into insurer portals, AI submission intake tackles that part.
Reading the results after two renewal cycles
Retention is the number that matters, so measure it properly before you change anything and again afterwards. A simple definition: policies renewed divided by policies that came up for renewal, excluding those where the insurer declined or the client's business closed. Track it by lapse-risk band, which tells you whether the calls are working: if high-risk clients now renew at nearly the rate of low-risk ones, the score is doing its job.
Alongside retention, watch the process measures: the share of questionnaires returned before day 60, the share of renewal reports sent by day 30, and the reasons clients give when they do leave. Ask every lapsed client one question on the follow-up call: "What would have made you stay?" Their answers are the best input for next year's rubric, and calculating AI ROI with a worked example shows how to turn retained premium into a figure you can compare with the cost of the tools.
The sum is usually short for a brokerage of this size. On 900 renewals, each percentage point of retention is nine policies kept. Say the average commission on those accounts is $600: one point is worth about $5,400 a year. Against that, six ChatGPT Business seats on annual billing come to $1,440 a year and Zapier Professional to about $240, so roughly $1,680 in tools. That's paid for by about three retained policies, before counting the handler hours saved. Use your own average commission; the point is that the break-even is small enough to test honestly over two cycles.
Further reads
- Where AI Saves Time in a Small Insurance Brokerage — Where renewals sit among a small brokerage's biggest time savings.
- What Does AI Cost a Small Insurance Brokerage in Year One? — Budget for the tools behind this workflow in year one.
- Can AI Answer Calls and Qualify Leads for an Insurance Agency? — Handle the inbound calls renewal reminders tend to generate.
- How Insurance Brokers Use AI to Handle Claims Enquiries — Claims experience is the biggest driver of renewal conversations.
- How to Automate Sales Follow-Ups With AI Without Being Pushy — Follow-up principles that stop reminders feeling pushy.
- How to Set a Baseline Before You Introduce AI — Measure retention properly before you change anything.
- How to Check Your Margins Product by Product With AI — Work out what each product really earns after discounts, shipping, fees and returns, with AI doing the sums in code and you checking three products by hand.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Zapier pricing and task-counting help pages; Make pricing page; HubSpot knowledge base on workflows by plan; ChatGPT Business and Claude Team plan pages.