AI Tools for Insurance Agencies: What to Use and What to Delegate

Two numbers from the same study tell the whole story.

In Liberty Mutual’s 2026 Independent Agency Growth Study, a survey of nearly 1,200 independent agency principals and staff, 65% of agents had used AI on the job in the past year — up from 37% a year earlier. Weekly use went from 18% to 41%.

And 14% said their agency had actually implemented an AI tool.

So the people in your agency have adopted AI. The agency hasn’t. That is not a contradiction, and it isn’t a failure of nerve. It is the most useful thing in the data, because the reason for the gap tells you exactly which work AI leaves on your desk.

Individual Adoption Is Easy. Operational Adoption Isn’t.

An account manager can have ChatGPT tidy their meeting notes this afternoon. Nobody’s permission is needed, nothing connects to anything, and if the output is wrong they notice and fix it.

Putting AI into the renewal cycle is a different kind of act. It has to reach the agency management system. Somebody has to decide what client information may be handed to which tool. Somebody has to own what happens when it is wrong, and somebody has to be there next Thursday when the sequence runs again.

The same study shows those harder measures barely moving while individual use nearly doubled. Principals who say they know enough about AI to make business decisions about it: 29%, up three points. Those who trust AI with client and business data: 22%, up three. Agencies with a well-defined AI use policy: 18%, up from 12%.

Individual use climbed 28 points in a year. Governance climbed six. That distance is the shape of the problem.

Where Do the Time Savings Actually Come From?

The same study asked where AI produced meaningful time savings. The ranking is familiar to anyone who has watched this happen in another industry:

• Summarizing meeting notes — 44%
• Generating marketing content — 43%
• Comparing policy details — 34%
• Automating routine tasks, such as data entry or document maintenance — 26%

Notes and marketing copy at the top. The actual operational automation at the bottom, at roughly half the rate.

And the ceiling shows up elsewhere in the same data. 8% of respondents said their agency uses AI to analyze customer data. 26% said they would not consider AI for taking a client’s initial contact information at all — a line agencies are drawing deliberately, not one they have failed to cross yet.

Four Hours a Week, and What It Doesn’t Buy

Agents using AI reported saving an average of four hours a week. More than half saved over two hours; 14% saved eight hours or more, which is a working day.

That is a real result and worth having. It is also worth being precise about what it is. Four hours a week is four hours returned to the person who was already doing the writing. It is not four hours of new capacity in the agency’s service operation, and it does not move a renewal through its runway, collect an updated schedule of values, or chase the carrier who has gone quiet.

Worth noting that contractors in the home services trades reported the same figure — just over four hours a week — in Housecall Pro’s 2025 survey of more than 400 operators. Two unrelated industries, the same number, and in both cases it is time given back to an individual rather than work taken out of the operation.

The Squeeze Behind All of This

Now the part that has nothing to do with AI, from the Big “I” and Future One 2026 Agency Universe Study, 1,376 respondents:

• The number of independent agencies fell from about 39,000 in 2024 to about 37,000 in 2026.
• Average staff rose from 8.2 to 9.9.
• Around 7 in 10 agencies reported higher personal lines revenue, and about 2 in 3 higher commercial lines revenue.
• The share reporting a revenue decrease fell from 12% to 8%.
• The top challenge agencies named was finding and screening job candidates with strong potential, at 45% — ahead of keeping up with AI at 43%.
• One in three agencies grew headcount over two years. Small agencies were the least likely to have added anyone; the largest were the most likely.

Read together: fewer agencies, each carrying more book, with revenue up and the binding constraint being people. And the agencies that added staff were the big ones.

That is the situation AI arrived into. It made the writing faster for individuals at agencies that were already short-handed, and left the service workload where it was.

The Cost of Standing Still Has Gone Up

One more set of findings, because they change the urgency. Zywave’s 2026 Broker Services Survey, published in July 2026 from more than 1,400 US employers, asked what would make a client consider changing brokers. The top three answers were slow response times, inconsistent communication, and the broker not acting as a strategic advisor.

Read that list again. Not one of those is a product problem. They are all service-capacity problems, and they are what a renewal calendar nobody has time for looks like from the client’s side of it. The same survey also found failure to leverage modern technology and AI-powered tools entering the top eight factors for the first time.

So the technology question is no longer only internal. It has become something a commercial client will weigh at renewal, which means the agencies with the least capacity to act on it are the ones carrying the most risk from not acting.

Where Does AI Earn Its Place in an Agency?

None of this argues for using less of it:

• Summarizing a long policy form or a carrier’s bulletin to orient someone before they read it properly.
• Marketing content — the newsletter, the renewal reminder campaign, the social post nobody has time to write.
• Comparing policy details side by side as a first pass, before a licensed producer reviews it.
• Meeting notes, which is the single highest-scoring use in the study and the least glamorous.
• Drafting routine client correspondence that a person then reads and sends.

The common thread is that in each case the output is text, and a person in your agency is the one who will act on it.

Where It Stops

Two limits, and they are different from each other.

The first is the licensed-work line, which agencies already understand well. An assistant gathers quote information; your producer prepares and presents the quote. An assistant does not bind, issue or change coverage, and does not explain, interpret or recommend coverage, limits or deductibles to a client. That line doesn’t move because a tool got better at drafting.

The second is quieter and it is the one the data keeps pointing at. Somebody has to decide what client information may go into which tool, and somebody has to read what comes back before it reaches a client or a carrier. With 18% of agencies holding a defined AI policy and 22% willing to trust AI with client data, agencies without a policy are making that call case by case, in the moment, by whoever is busiest. That judgment sits with your agency rather than with a tool.

There is a third limit that sits underneath both, and it has nothing to do with drafting. A workflow set up months ago and trusted since — the renewal reminder, the certificate request routing, the sync that keeps the AMS current — stops without announcing it. Automation does not fail loudly, and because it was built to run unattended, nobody is watching the point where it stopped. In an agency the consequence is specific: a renewal nobody worked, found at the deadline rather than ninety days out.

The harder version is a workflow still running that no longer does what it was built to do. Nothing errors, the reminders still send, and the service the client was supposed to receive quietly stopped arriving. That takes somebody close enough to the service calendar to notice a renewal queue that has not changed, and to know that is wrong.

What Should You Hand to a Person?

The work AI did not take is specific, and it fills the service calendar: The service-account workload is set out stage by stage in policy checking, endorsements and renewal prep.

• The renewal runway: what happens at 90, 60, 45, 30 and 10 days out, and the reminder that goes before the deadline rather than after
• Certificates of insurance prepared from existing policy data, and the contractor who needs one by Friday
• Collecting updated exposure information — payroll figures, vehicle schedules, updated values — and getting it to the carrier
• Endorsement paperwork routed and tracked to completion
• Chasing the carrier submission that has gone quiet, and the client who hasn’t returned the application
• AMS hygiene: activities and notes logged after client calls, the same client merged out of three records, renewals that are actually dated correctly
• Marketing executed on a schedule rather than in the gaps between renewals

A VRA assistant administers that back office under your agency’s direction.

What This Has to Do With the Hiring Number

Return to the 45%. The hardest thing agencies reported wasn’t AI strategy — it was finding and screening candidates with strong potential, and small agencies were the least likely to have added anyone.

That is the gap VRA exists in. Assistants come through screening, a mandatory four-week instructor-led program, and a final evaluation before they touch client work. 1,800 applications produced 16 graduating assistants, or 0.89%. The screening has already happened by the time you meet someone.

And a hire is a salary, plus payroll taxes, benefits and equipment, carried whether the quarter is good or not. If the book is clearly growing and the money for another salary isn’t there yet, the service work stays where it is and the things that would grow the agency wait behind the things that keep this week running. You shouldn’t have to just endure that. A virtual assistant or your first employee goes through that comparison in detail.

An Assistant Who Already Uses AI

The choice was never AI or a person. It is a person who uses AI.

Each VRA assistant completes the Essential Toolkit before their first client interview, and Claude and ChatGPT are part of it, alongside Canva, a CRM, Zapier and Asana. Assistants who need deeper AI skills for a client’s business can go further, through a dedicated AI track.

So what reaches you is the finished result, handled to your standards, rather than a draft you still have to act on, because somebody inside your operation now owns the part that was never going to run itself.

More on how it works on our insurance agency virtual assistant page, and on the wider question in should you just use AI instead of hiring a virtual assistant?

What Does It Cost?

A VRA assistant is $1,500 a month full-time, or $800 a month part-time for 20 hours a week. One flat rate, with no setup fee, no overage charges and no upcharge for a bilingual assistant. Increased coverage is available at any time.

AI in Insurance Agencies: Common Questions

What happens when an agency automation quietly stops working?

It costs a renewal before anyone notices. Reminder workflows, certificate routing and AMS syncs fail without announcing it, and because they were set up to run unattended, nobody is watching the point where they stopped. The harder case is the workflow still running that no longer does its job, where nothing errors and the service simply stops arriving. Catching either one takes a person working the service calendar daily, close enough to see that a renewal queue has not moved.

How are insurance agencies actually using AI?

Liberty Mutual’s 2026 Independent Agency Growth Study found the biggest reported time savings in summarizing meeting notes (44%), generating marketing content (43%), comparing policy details (34%) and automating routine tasks such as data entry (26%). The Big “I” 2026 Agency Universe Study found 46% of agencies reporting some AI use, up from 15% in 2024, with marketing content generation, coverage form analysis and contract reviews the top three uses. The two studies measure different things and are not in conflict: the Big “I” figure counts any AI use at an agency, including a staff member using ChatGPT on their own, while Liberty Mutual’s 14% counts agencies that have implemented an AI tool. In both, the heaviest use is on producing and summarizing text.

Why do individual agents use AI more than their agencies do?

Because the two are different acts. In the Liberty Mutual study, 65% of agents had used AI in the past year while 14% said their agency had implemented an AI tool. An individual can adopt a drafting tool alone. Putting AI into a service workflow means reaching into the agency management system, deciding what client data may be used, and having somebody own the result — which is work that needs a person assigned to it.

Can a virtual assistant do insurance work without a license?

Administrative work, yes; licensed work, no. A VRA assistant handles certificates of insurance, renewal preparation and reminders, AMS data entry, endorsement paperwork, collecting quote information and applications, scheduling and routine client service email. Quoting, selling, binding and advising a client on coverage, limits or deductibles stay with your licensed staff.

How much time does AI actually save an insurance agent?

An average of four hours a week among AI users, according to the 2026 Liberty Mutual study. More than half saved over two hours and 14% saved eight hours or more. That is time returned to the individual doing the writing, not added capacity in the agency’s service operation.

What should an agency delegate to an assistant instead of AI?

The renewal runway and its reminders, certificates prepared from existing policy data, collecting updated exposure information and getting it to the carrier, endorsement paperwork tracked to completion, chasing quiet carrier submissions and unreturned applications, AMS data hygiene, and marketing executed on a schedule.

Do VRA assistants use AI?

Yes. Claude and ChatGPT are part of the Essential Toolkit each assistant completes before their first client interview, and there is further AI training for assistants whose clients need it.

How much does an insurance agency virtual assistant cost?

$1,500 a month for full-time at 40 hours a week, or $800 a month part-time for 20 hours. One flat rate, with no setup fee and no overage charges, and increased coverage is available.