A Midwest insurance agency cut call handle time from 11 to 6 minutes with one AI agent. Here's exactly how it worked.
Article text
The office manager pulled me aside during our second visit and pointed at the phone system dashboard. It was 2:47 PM on a Tuesday. Eleven calls in queue. Two clients already on hold for over six minutes. One rep was apologizing to a caller while scrolling through a PDF. Another had three browser tabs open, hunting for a coverage summary that should have taken ten seconds to find.
"We lost three clients last month," she said. "Not because we gave bad advice. Because they couldn't get through to us fast enough."
This was a 35-person insurance agency in the Midwest. Personal lines, commercial lines, benefits. Good team, solid book of business, growing steadily for eight years. And they were bleeding clients because their customer service workflow hadn't been updated since they were a 15-person shop.
If your team spends more time searching for information than actually helping the person on the phone, that's the exact pattern we help fix . It takes less than you think.
The Problem Wasn't the People
The easy diagnosis would have been "not enough staff." The agency owner had already considered hiring two more CSRs. At $42K per rep plus benefits, that's roughly $110K in new annual overhead before the first call gets answered.
But we don't start with solutions. We start with mapping the actual process , step by step, and figuring out where the friction really lives.
Here's what we found when we timed a typical inbound service call:
Caller identifies themselves
(15-30 seconds)
Rep searches for the client in the agency management system
(45-90 seconds)
Rep opens the policy file and navigates to the right coverage section
(60-120 seconds)
Rep cross-references a second document
(endorsement, certificate, or claims history) (45-90 seconds)
Rep answers the actual question
(2-4 minutes)
Steps 2 through 4 are pure information retrieval. The rep isn't advising, selling, or solving anything. They're hunting. On average, that hunt took 4 minutes and 12 seconds per call.
With 85-100 inbound calls per day across six reps, that's roughly 350-420 minutes of daily labor spent on information retrieval alone. Nearly 7 hours of collective work every day that added zero value to the client experience.
The average handle time: 11 minutes per call.
The Diagnosis
The bottleneck wasn't headcount, phone systems, or training. It was the gap between when a call starts and when the rep has enough context to actually help.
Three separate systems held the information each rep needed: the agency management platform, a document storage system for policy PDFs, and a separate portal for claims history. None of them talked to each other. Every call required manual lookups across all three.
We also found something the agency hadn't tracked: error rates. When reps are rushing through lookups while a client is waiting on the line, mistakes happen. They were quoting wrong deductibles about 8% of the time, misidentifying coverage limits in roughly 1 out of 12 calls. Most of these errors got caught before they caused real damage, but a few had turned into E&O close calls.
The solution wasn't "replace the reps with a chatbot." The clients wanted to talk to a human. They just wanted that human to already know who they were.
What We Built
One agent. Not a chatbot. Not a phone tree. A background lookup tool that works alongside the rep, not instead of them.
Here's how it works:
When a call comes in, the agent matches the incoming number to a client record. By the time the rep picks up the phone, a single-screen summary is already waiting: client name, active policies, coverage highlights, recent claims, upcoming renewals, and any open service requests.
The rep greets the client. The information is already there. No tab-switching, no PDF scrolling, no cross-referencing.
We integrated it with the agency's existing management system and document storage through their APIs. No rip-and-replace. The reps kept using the same tools they already knew. The agent just pre-loaded the context.
The build took 3 weeks from kickoff to live deployment. One week of process mapping and integration scoping. One week of building and testing the lookup agent. One week of running it alongside the existing workflow so reps could get comfortable and we could catch edge cases.
The whole thing runs on less than $200 per month in compute costs.
The Results
We measured everything against the same metrics from the diagnostic phase. Here's what changed in the first 60 days:
Handle time dropped from 11 minutes to under 6.
The 4-minute information hunt collapsed to about 15 seconds. Reps went from five steps to two: greet the client, answer the question.
The call queue cleared by 2 PM instead of backing up at 3 PM.
Same six reps, same call volume. They just moved through calls nearly twice as fast.
Client retention improved 14% in the first 60 days.
The agency had been losing 2-3 clients per month to competitors who answered faster. In the two months after deployment, they lost zero clients to service complaints.
Quoting errors dropped by roughly 70%.
When reps aren't rushing through manual lookups under pressure, they get the numbers right. The 8% error rate on deductible quotes fell to under 2.5%.
The agency didn't hire those two extra CSRs.
That's $110K in annual overhead they didn't take on. Against the build cost and $2,400/year in operating costs, the math speaks for itself.
What This Cost vs. What It Saved
The numbers on this one are straightforward:
Build cost:
Mid four figures (one-time)
Monthly operating cost:
Under $200
Annual operating cost:
Under $2,400
Avoided hiring cost:
~$110K/year (two CSRs not hired)
Revenue protected:
Three clients per month were leaving. Average client lifetime value at this agency was around $4,700. That's roughly $170K in protected annual revenue.
Error reduction:
Fewer E&O exposure incidents, fewer correction calls, fewer "I need to call you back" moments
As Five9's recent study showed , customer service is where AI is delivering some of the clearest ROI across industries. The enterprise numbers are bigger, but the pattern is identical at the SMB level: reduce information retrieval time, and everything downstream improves.
The Takeaway
This wasn't an AI transformation. It was a plumbing job. The agency had good people, good clients, and a good business. They had one workflow that hadn't scaled, and it was costing them clients they'd spent years building relationships with.
The agent we built does one thing: it gets the right information in front of the right person before they need to ask for it. That's it. No machine learning magic. No "AI-powered insights." Just a lookup tool that's faster than a human clicking through three systems.
If your team is drowning in calls, the answer usually isn't more bodies. It's removing the friction that makes each call take twice as long as it should. Sometimes that fix costs less per month than your office coffee budget.
Dealing with the same pattern? Send us a message . We'll tell you in 30 minutes whether it's worth building something.