Insurance brokers: never let a quote request go cold
A quote request without a quick reply goes to a competitor. Here is the sequence that acknowledges it, asks for what is missing and prepares the file, without ever advising in your place.

A quote request that waits goes cold fast
Short answer: to stop quote requests from going cold, a brokerage should reply quickly with a first useful message, ask for the missing details needed to price the risk, then hand the complete file to the right person. AI can prepare the first two steps. The recommendation and the call stay with you.
Someone asking for an insurance quote almost never contacts a single firm. They fill in a form with you, another with a comparison site, and perhaps write to a competitor. The first to answer with something concrete gets the edge. The next one speaks to a person who is already in a conversation.
In a brokerage these requests arrive through four doors: the website form, the shared inbox, the phone and, more and more, messaging apps. They also arrive when nobody is free: evenings, weekends, during a client meeting. The work starts with sorting, not advice. That sorting is what we propose to hand to the machine, along the lines described on our page on qualifying inbound requests and in our work for insurance brokers.
What delay costs, with the figures available
To our knowledge there is no independent study measuring average response time in brokerage. What exists are figures published by messaging-tool vendors, so read them as vendor claims.
- A messaging vendor reports a US health-insurance broker whose first-response time reportedly fell from about 15 minutes to about one minute, with 87.5% of conversations handled by AI agents (case study published by respond.io).
- Another vendor, Trengo, claims a 90-second prequalification against two days of manual review, and 70% fewer abandoned quotes thanks to automated follow-ups. The page cites no independent source.
- A blog post from a French provider in this niche describes a comparison site answering within minutes where the broker answers a day or three later, and a prospect who is “sometimes already gone”.
Keep the direction rather than the values: the gap plays out in minutes or hours, not weeks. Measure your own before trusting anyone. Take the last twenty requests, note the arrival time and the time of the first real reply (not the website’s automatic receipt). That median is your starting point.
The four places a request gets lost
- The shared inbox. Everyone sees it, so everyone assumes someone else has it. A request can stay “seen” for a whole day.
- Evenings and weekends. The request arrives at 7:40 pm on a Friday. It is read on Monday at 8:30 am. In between, the prospect has had an answer elsewhere.
- The incomplete file. The current policy’s expiry date, the driver’s claims history, the date of birth or the company’s turnover is missing. You have to write back, and that first exchange costs two days.
- Slow pricing. Even with all the data, building the comparison takes time. If nothing tells the client they are being handled, they assume they are not.
The first two are solved by a useful acknowledgement. The third is solved by asking the right questions straight away. The fourth calls for planning, not automation: a clear delivery promise is often enough to keep the prospect.
The sequence, step by step
Definition: qualifying a quote request is the step that turns a raw message (“I’m looking for insurance for my car”) into a usable file: type of risk, expiry date of the current policy, data needed for pricing, urgency and preferred way to be called back. It is the most repetitive part, and the one where delay costs the most.
- Receipt and identification. Whatever the channel, the request is logged in one place with its arrival time. The system checks whether the person is already in your database and whether a file exists.
- Sorting by line of business. Motor, home, health, life and pensions, commercial: each line calls for different questions. The AI sorts the request and flags those it cannot classify.
- First message within minutes. A short message in your name that confirms the request is being handled, states an honest deadline (“a proposal by Thursday noon”) and asks the three or four questions that are missing. It starts from a template you wrote.
- Collecting details and documents. The prospect’s answers are read, filed and checked against the list of data required for that line. If something is missing, one reminder goes out, not ten.
- A file ready to price. Once the file is complete, a one-page sheet reaches the designated person: need, current policy and expiry, pricing data, points of attention, conversation history.
- You take over. You call back, advise, price and send. The system makes neither the recommendation nor the sending of the proposal.
- Watching the silence. If a proposal gets no answer, a reminder to the responsible person appears on the planned date. For the follow-up itself, see our article on following up on a quote without being pushy.
What AI prepares, what stays with you
| Step | AI prepares | You decide |
|---|---|---|
| Acknowledgement | Message in your name, stated deadline | The template and a deadline you really keep |
| Sorting by line | Classification and priority | Out-of-scope or sensitive requests |
| Missing questions | List adapted to the line | Questions never asked in writing |
| Case sheet | Summary, data filed, documents | Checking free-text fields |
| Pricing and advice | Nothing decisive | Everything: cover choice, recommendation, sending |
The reliability limit sits in free-text fields. The public sources we read say extraction is good on structured fields (dates, plates, addresses) and more variable on a free description of a need. Our rule: any field that changes the price is checked by a person before pricing.
A calculation to run on your own numbers
The calculation needs no studio figure. It needs three numbers from your firm. Take illustrative values, to be replaced with yours: 40 quote requests a month, 25% arriving outside office hours, and a third of those lost for lack of a quick answer.
That gives 10 out-of-hours requests, about 3 lost a month, roughly thirty a year. Multiply by the average annual premium of a policy in your book and by your commission rate: you get the yearly shortfall due to delay alone. The three assumptions are arbitrary, and that is the point: change them. If, for you, the figure is two lost requests a year, the topic is not a priority and you should start elsewhere.
Writing the first message: what it must contain
The first message does most of the work. It fits in five lines and follows a fixed order: one sentence confirming the request is being handled, the name of the person handling it, the day and time by which the prospect will have a proposal, the three or four missing questions, and one sentence on how to be called back. Nothing else: no brochure, no firm presentation, no price promise.
Two mistakes come up again and again. The first is announcing a deadline the firm does not keep: a missed deadline does more damage than a long one that is respected. The second is asking ten questions: the prospect gives up, or answers halfway. Keep what changes the price and leave the rest for the call. One template per line of business, reviewed once a quarter, is enough.
When not to automate this
- If you receive few requests. Five requests a month are best handled by hand. A system only pays off at a volume where sorting takes your time or loses files.
- If your clients phone. For business or wealth clients, the phone remains the channel of trust. An automatic message can bother more than reassure. In that case, automate the case sheet and the internal reminder, not the reply to the client.
- If the risks are complex. An industrial risk, a bespoke cover: the first question is already advice. Do not hand it to a model.
- If your sequence is not written down. Automating practices everyone runs their own way freezes the confusion. Write the questions per line and the deadline you can keep first.
Client data and the framework to check
A quote request contains personal data: identity, vehicle, family situation, sometimes health. Before plugging in any tool, check where the data is processed, who can read the sheets and whether every sending is logged. Our article on the Swiss data protection act and data sent to an AI model sets out the questions to ask. This article is not legal advice: validate the framework with your firm’s compliance lead.
Where to start this week
List your last twenty quote requests. For each, note the arrival time, the time of the first real reply and what was missing from the file. You get three things: your real delay, your quiet hours and the list of questions you always ask again. Those notes are the specification for the first message. To place your case before deciding, the five-question express diagnostic takes a few minutes. And if contract expiries are your other leak, also read renewals and expiry dates without a spreadsheet.
Frequently asked questions
What response time should I aim for on an insurance quote request?
There is no independent standard. Tool vendors cite replies within minutes. First measure your current median on your last twenty requests, then aim for a useful first reply the same day and a clear deadline for the proposal.
Can AI reply to a prospect on its own?
It can send an acknowledgement in your name and ask the questions missing for pricing. It does not make the recommendation, choose the covers or send the proposal. Someone at the firm approves the message template and takes over the file.
What should I do with requests received in the evening or at the weekend?
Send a message straight away confirming the request is being handled and stating the day you will reply, then prepare the case sheet for the first working slot. The prospect knows they are being dealt with, and the firm does not have to answer live.
Which details should the first message ask for?
It depends on the line: current policy expiry, vehicle or home data, date of birth, the company’s activity. Write a short list per line, three or four questions, and keep the rest for the call.
Is this automation suitable for a small brokerage?
It pays off when sorting requests takes time or loses files. With a few requests a month, handling them by hand gives a better result. The diagnostic lets you check that on your own numbers.


