AI Receptionists
Nobody Ever Forgets a Certificate Request: How Service Work Quietly Outranks New Business at Insurance Agencies
Service requests arrive with a name, a document, and a deadline. New quote requests arrive with none of those, so they lose the queue every day without anyone deciding they should. Six counts an agency owner can run this week to find out how much that costs.
by Jerrod Anthraper
A contractor calls at 9:15 needing a certificate of insurance emailed to a general contractor before his crew can get on site at ten. Your account manager drops everything, pulls it, sends it, confirms receipt. Twenty minutes later a woman calls about coverage on a rental property she is closing on next month. She goes on the callback list.
Nobody made a decision there. No manager ranked those two calls. The queue ranked them, and the queue will rank them the same way tomorrow.
The leak: service work has deadlines, new business does not
Every service request that hits your agency arrives with three things attached: a named person, a specific document, and a deadline that is usually today. The COI, the ID card, the endorsement, the lienholder change, the billing question about a draft that hit wrong.
A new quote request arrives with none of that. No deadline, no document, no consequence at 4:45 if it did not happen. So it loses. It loses every single time, to every single service request, all day, without anyone noticing they are choosing.
Now put a number against it. Above a 32% win rate, achieved with roughly 27% of the work of a human sales team, is the benchmark our framework runs at. Read that second figure as what it is: a statement about how much of intake is repeatable process rather than judgment.
Why it happens: your best people are correctly prioritizing the wrong thing
This is not a training problem, and it is definitely not a caring problem.
Your CSRs and account managers are graded, formally or otherwise, on service. Retention shows up in renewals. A missed COI shows up as an angry contractor and a general contractor who remembers. A missed quote request shows up as nothing at all, because the person who called simply bought from someone else and you never learn their name.
That asymmetry is the whole mechanism. One failure is loud and immediate. The other is silent and permanent. Any rational person under time pressure handles the loud one first, which means your new-business queue is being managed by whoever happens to have a free minute, which is nobody, most days, between about nine and four.
The diagnostic: six things to count this week
Run these counts before you change anything. They take a few hours total and they will tell you whether this is your leak or not.
1. Log every inbound call for three days by type
One tally sheet, four columns: service, billing, claims, new business. Every call gets one mark. No names, no notes.
Most agency owners guess that new business is fifteen or twenty percent of call volume. Most are surprised when they see the real split, and more surprised by what time of day the new-business calls actually land.
2. Time-stamp new-business calls from arrival to human contact
For those same three days, write down when each new-business inquiry arrived and when someone actually spoke with that person. Not when you left a message. When you connected.
Then sort. You are looking for the shape of the distribution, not the average. If a handful sat over four hours, that tail is where your money is.
3. Count how many quote requests died without a second attempt
Pull last month. For every new-business inquiry that did not become a quote, count how many follow-up attempts were made after the first one.
In most agencies the honest answer is one attempt, or one attempt plus a voicemail. Not because anyone decided one was enough, but because there is no list anywhere that says who still needs a second call.
4. Ask your team one question, and mean it
Ask each person on the phones: "When a new quote call comes in while you are in the middle of a service task, what are you supposed to do?"
If you get four different answers, you have your finding. If you get four versions of "get to it when I can," you have it in writing. Neither answer is a performance problem. Both are a design problem, and design problems are the ones you can actually fix.
5. Check the hours nobody is assigned to
Finally, look at what happens to an inquiry that arrives at 5:40 p.m., on a Saturday, or during your Tuesday staff meeting. Not what the voicemail says. What actually happens to that person, and how long it takes.
Those hours are unstaffed by definition. The question is whether they are unstaffed by choice, and whether anyone has ever counted what arrives in them. Across service industries, 40% to 60% of inbound leads land outside business hours and are never answered at all, which means the queue problem you have during the day is only part of the picture.
6. Put a dollar figure on it before you do anything else
Take the number of new-business inquiries from last month that never became a quote. Multiply by your typical bound-policy commission, then by whatever close rate you would honestly expect if a producer had actually reached every one of them.
Be conservative. Use a close rate you would defend in front of your producers. The point of this step is not to build a case for change — it is to find out whether the leak is worth a Tuesday afternoon of your attention or whether it is genuinely small and you should go work on something else. Plenty of agencies run this and find the number is modest. Those agencies just saved themselves a purchase.
The proof
Our framework benchmark holds above a 32% win rate while requiring roughly 27% of the work a human sales team would put in — near parity on outcome, at a fraction of the labor. That is not a claim that people are replaceable. It is a claim about which part of the job is repeatable.
The first ten minutes of a new-business call are almost entirely repeatable: who are you, what are we insuring, what do you have now, when does it renew, when can a producer talk to you. That is the part the queue keeps losing. The judgment part, the part your licensed producer is actually good at, only starts after that.
Worth being clear about the limit, since this is an industry with real licensing rules. Nothing above suggests automating advice, recommending coverage, or binding anything. Intake is not counsel. Separating the two is what lets the licensed conversation happen at all, because right now a fair number of those conversations are simply not happening.
The next step, and it is a small one
Every step above is manual on purpose. Run it first so you know your own numbers before anyone sells you anything.
If it turns out that new-business calls are losing the queue, that is exactly the gap our AI agent, James, was built for. He takes the new-business inquiry the moment it arrives, at 9:15 or at 9:15 p.m., runs the intake questions, qualifies it, and puts a booked appointment on your producer's calendar. Your account managers keep doing service work without having to referee between a contractor who needs a COI in twenty minutes and a stranger who wants a quote.
Start with the three-day tally sheet. If new business turns out to be a meaningful share of your call volume and most of it is going to a callback list, that is a number worth a conversation.