Book a call

Case studyExample business, real tools, estimated numbers

Seven AI agents run an insurance agency's busiest night

In this case study, seven AI agents run the night of a small independent insurance agency in Lancaster, Pennsylvania. A family closes on a house Friday and their lender wants proof of insurance by noon. Overnight, the agents answer the call, read the lender's email, prefill the quote in EZLynx and lay five carrier quotes side by side. At 8:30 AM the licensed agent advises and binds, and the lender has its proof at 9:20 AM.

The video loads from YouTube only when you press play.

Business
Ashford & Price Insurance, an independent agency in Lancaster, Pennsylvania
Size
The owner, who is the licensed agent, a licensed account manager, about 2,100 policies
The night
Monday 10:58 PM to Tuesday 9:20 AM, closing on Friday
Agents
Seven: Otto, Intake, Docs, Rater, Brief, Proof and Guard
Tools
EZLynx Outlook Claude
Build time (estimate)
About three weeks, then the AI officer plan
12of 12
after-hours calls and texts answered (from voicemail)
One night
9:20AM
proof of insurance to the lender (from late morning, near the noon cut-off)
Estimated
0
policies bound or proofs sent without the licensed agent
By design
Cover of the guide: The AI Insurance Agency Night Shift

Free guide · PDF · 79 pages

The AI Insurance Agency Night Shift

Implementation guide for independent agencies: the seven agents with their build specs, the rules, the rollout and the templates.

View the guide Download PDF

The situation

There are about 37,000 independent insurance agencies in the United States, with fewer than ten people on average. In most of them, the owner is also the licensed agent who advises clients, and the busiest moments rarely arrive during office hours.

Home buying is one of those moments. Existing homes were selling at an annual pace of 3.98 million in August 2026, and when a buyer has a mortgage, the lender wants proof of insurance before closing: the policy details, its mortgagee clause and the loan number. Lenders set their own cut-off. Miss it, and the closing can move.

In this demo, that request lands at 10:58 PM on a Monday. Emily and Josh Carter close on their first home in Lititz on Friday. Their lender, Lakeside Mortgage, wants the evidence of insurance by noon on Tuesday. The agency is closed. Tom Ashford, the owner and licensed agent, is asleep.

How the night used to go

At a small agency, a request like this usually starts as a voicemail. The next morning someone listens to it, calls the buyers back, asks them to forward the lender's email, types the house into the rater, waits for the carriers, and only then can the agent talk to the family about coverage. The proof goes out late in the morning, close to the cut-off, often after a second call from the lender's processor.

None of that work is the agent's real job. The real job is the twenty minutes on the phone, explaining the options and recommending the coverage. Everything around it is gathering, typing and chasing, and it all happens in the hours when the agent should be advising.

The agents

Seven agents, each with one job and a written rulebook, all working inside the agency's own tools. Otto leads the night; the other six are specialists.

OttoLead

Runs the night: hands out the work, keeps the log, and gives Tom a two-minute summary at 7:40 AM.

Never makes a decision for the agency.
Intake

Answers every call and text, says it is the agency's AI assistant, and sends a secure link for the lender's email and the inspection report.

Never recommends coverage or quotes a final price.
Docs

Reads the lender's email and puts the mortgagee clause, the loan number and the closing date on the application.

Never changes what the lender asked for.
Rater

Prefills the house in EZLynx from property data, including the replacement cost estimate, and rates it with six carriers at once.

Never picks the carrier.
Brief

Lays the quotes side by side for the agent, with the questions only he can answer, like the 17-year-old roof.

Never marks a quote as the one to buy.
Proof

Once the agent binds, sends the carrier's evidence of insurance to the lender, and tracks every closing until the policy is bound.

Never sends anything before the licensed agent signs off.
Guard

Checks every message before it leaves: no advice, no price, nothing bound, AI disclosed, texts only with consent.

Never lets a message through that breaks a rule.

Step through the night

Press play, or tap any time, to follow the request from the lender's email to the proof.

Monday night to Tuesday morning
10:58 PM
Docs

The buyers forward the lender's email to the agency.

Lakeside MortgageEvidence of insurance with our mortgagee clause and loan #4418-2207 needed by 12:00 PM Tuesday, or Friday's closing moves.
1 / 9

The brief Tom read at 7:40 AM

Annual premium for the same house, Coverage A $412,000, $1,000 deductible. Carrier names are hidden on purpose.

CarrierAnnual premiumNote for Tom
Carrier A$1,412Ask: how is a roof claim settled?
Carrier B$1,486
Carrier C$1,538
Carrier D$1,620
Carrier E$1,745
Carrier FNone yetWants a roof inspection first
 The licensed agent recommends; the brief only lays out the options.

The stack

No new software for the team. The agents work inside what the agency already pays for, and the person who advises stays the same.

EZLynx, the rater Outlook, the inbox Claude, the agents
  • EZLynx already prefills home details from property data and rates many carriers from one entry. Rater fills it in from what Intake and Docs gathered, so nobody types the house twice.
  • The carrier's portal is where the licensed agent binds. No agent has the authority to bind, so none is given access to do it.
  • Outlook is where the lender's email lands. Docs reads it there and files it with the client.
  • The phone line forwards after-hours calls and texts to Intake. The callers hear that they are talking to an AI.

Before and after

MomentBefore (typical)With agents
After-hours callVoicemail, called back the next morningAnswered at 11:02 PM, AI disclosed, details taken
Lender's emailRead the next morning and retypedClause, loan number and closing date on the application by 11:20 PM
QuotesEntered by hand after the call-backSix carriers rated overnight, five quotes in the brief
The agent's morningGathering and typingAdvising: an 8:30 AM call with everything in hand
Proof to the lenderLate morning, close to the noon cut-off9:20 AM, right after the agent binds
E&O fileNotes written when there is timeEvery offer, answer and decision logged
 Before: a typical small-agency morning, estimated. With agents: the night in this case study.

The value, and how we estimated it

Counted in hours and outcomes, with each assumption written out so you can redo the math with your own agency's numbers.

ValueEstimateHow we estimated it
Team hours backabout 300 a yearAbout 6 hours a week of after-hours callbacks, retyping and quote entry move to agents, over 50 working weeks.
After-hours calls answeredall of themCalls and texts forward to Intake instead of voicemail.
From binding to proofunder 20 minutesProof prepares the evidence while the agent binds and sends it on sign-off: 9:05 to 9:20 AM in this case study.
Closings followedevery one, until boundProof tracks each closing until the policy is bound. Coverage never placed is the most common claim against agencies.
 Estimates, with every assumption shown.

How a build like this would run

Estimated for an agency of two to ten people that already uses EZLynx and Outlook.

  1. Week 1

    Map

    Listen to a week of after-hours calls, read the lender emails, list the carriers set up in EZLynx. Agree on the result, for example: every closing request answered the same night, and proof sent within an hour of binding.

  2. Weeks 2 to 3

    Build

    Intake on the after-hours line first, then Docs and Rater. Guard's rules are written with the owner, line by line. The team tests on ten real files before anything goes live.

  3. Day 30

    Prove

    The result is measured from the agency's own phone and email logs. The second half of the setup is due only then.

  4. Every month

    Stay

    A monthly report, updates when a carrier changes its forms, and one improvement, such as renewal reminders by text for clients who said yes.

What stays with the people

The agents gather, prefill and draft. The licensed agent advises, binds and signs. These lines are written into Guard's rulebook.

No coverage advice from the AIIn Pennsylvania, advising a buyer on a policy's terms takes a producer license (40 P.S. § 310.1).
No final price, nothing boundOnly a licensed agent appointed by the carrier binds.
No proof before the agent signs offThe lender wants evidence from the insurer (Fannie Mae Selling Guide B7-3-07).
The AI says it is AI, and texts only people who said yesThe FCC treats AI voices as artificial under the TCPA (2024 ruling).
Every step is logged"Coverage not procured" is the most common claim against agencies, about 30% (Swiss Re, 2020 data).

What's in the guide

The full implementation guide behind this case study, 79 pages, free to read and to download. Built so a agency owner, or whoever builds the system with them, can follow it.

  1. The business reality: the numbers, the daily workflow and the tools already in place, with sources.
  2. One closing night at Ashford & Price, minute by minute, with the runbook.
  3. The seven agents: a job card and a build spec for each, and the knowledge they share.
  4. The rules: producer licensing, binding, evidence of insurance, the NAIC AI bulletin, calls and texts, privacy, rebating, E&O.
  5. Build and run it: the stack, the data flow, the 30-60-90 day rollout, supervision and the numbers to track.
  6. Templates and appendices: the Claude skill and sixteen agency files, scripts, a glossary and every source.

View the guideDownload PDF

Where the numbers come from

The rules and tools are real. The agency, the people, the lender, the loan number and every premium are invented for this case study, and the before and after are estimates.

Read the video transcript

Narrator: I built seven AI agents that run an insurance agency. Here's its busiest night. Eleven PM. A family closes on a house Friday. Their lender wants proof of insurance by noon, or the closing moves. Homes are selling at almost four million a year, and lenders want this proof before closing. And the blueprint is free.

Narrator: Otto runs the agency's night, with six specialists: Intake, Docs, Rater, Brief, Proof, and Guard. All inside the agency's own tools.

Intake (AI): Ashford and Price Insurance, this is the agency's AI assistant. How can I help?

Emily: Hi. We close on our house Friday, and the lender needs proof of insurance by noon tomorrow!

Intake (AI): Then let's start tonight. Can I text you a secure link for the lender's email and your inspection report?

Emily: Yes, please. Which policy should we get? And how much will it be?

Intake (AI): I can't recommend coverage or quote a final price. Tom, your licensed agent, will go over your options at eight thirty.

Narrator: Intake answers every call and text, and says it's AI. Docs reads the lender's email: the clause, the loan number, the closing date. Rater prefills the house in EZLynx, and runs it past six carriers at once. And Brief lays the quotes side by side for the agent, with the questions only he can answer.

Narrator: But nothing that commits the agency happens overnight. No coverage advice, and no final price. Nothing gets bound. And no proof goes to the lender until a licensed agent signs off. In Pennsylvania, advising a buyer on coverage takes a license.

Brief (AI): Morning, Marcus. The Carters close Friday. Five carriers quoted their house overnight, and one wants a roof inspection first.

Marcus: How old's the roof?

Brief (AI): Seventeen years. And the lender's clause and loan number are already on the application.

Marcus: Nice. Then Tom can bind this morning.

Tom: Morning, Otto. How was the night?

Otto (AI): Twelve calls and texts. The Carters close Friday, and their lender wants proof by noon. Five quotes are in your brief.

Tom: Did they ask which one to buy?

Otto (AI): They did. I said you'd recommend the coverage. Nothing is bound, and nothing went to the lender.

Tom: Good. I'll call them at eight thirty.

Narrator: At eight thirty, Tom calls the Carters, brief in hand, and recommends the coverage. He binds it in the carrier's portal. Then Proof sends the carrier's evidence to the lender, before ten.

Narrator: And Proof tracks every closing until the policy is bound. Because the most common claim against agencies is coverage that never got placed.

Narrator: The AI gathers, prefills and drafts. The licensed agent advises, binds and signs. Comment POLICY, and the blueprint is yours.

For insurance agencies

Want this for your agency?

In 45 minutes we look at one job that keeps your team late, and tell you if an agent can do it inside the tools you already use.

Get a free 45-minute map

More case studies

A law firm's intake, all night7 agents · Law firm A one-man HVAC shop's back office7 agents · Home services