Independent agents have taken up AI largely on their own. Many use it to draft emails, summarize policy forms and write marketing copy. Far fewer agencies have built it into their own systems.

In Liberty Mutual's 2026 Independent Agency Growth Study, a survey of nearly 1,200 US agency principals and staff, 65 percent of agents said they had used AI for work in the past year. Only 14 percent said their agency had actually put an AI tool in place.

Agency management system vendors, carriers and insurtechs are all selling AI products to agencies, alongside free general-purpose chatbots. That leaves agency owners deciding whether staff will use another tool, and whether the agency can absorb the change.

"A change management problem"

Eric Purdy of Rival Insurance Technology, an agency software company, told Insurance Business in a recent interview that brokers spend much of the day pulling together information from separate places: the management system, email, documents, quoting tools and carrier portals.

Swapping out a system an agency has used for decades raises fair questions about data, retraining and integrations, he said, calling the challenge "a change management problem." He was also skeptical of vendors that add a chatbot to an old platform and call the result AI-native.

Read next: Independent agency revenue rose at three in four firms as AI adoption tripled, Big I study finds

Agents adopt AI faster than agencies

Liberty Mutual's figures, which it has tracked for three years, show the gap widening. Individual use jumped between its 2025 and 2026 surveys; the measures of agency oversight barely moved.

Agents who use AI say it saves them an average of four hours a week. The share whose agency has rolled out an AI tool more than doubled to 14 percent, though Liberty Mutual did not publish the exact  2025 starting figure to make that calculation. It attributes the slower agency uptake partly to resources: anyone can try a free chatbot, but deploying a tool across an agency is a bigger job, and larger agencies are more likely to have done it.

Research from the Big "I" shows a similar pattern. Its 2026 Agency Universe Study put agency-level AI use at 46 percent, up from 15 percent in 2024, and "keeping up with the latest in AI" entered the list of top agency challenges at 43 percent, just behind hiring. Its separate Agents Council for Technology survey found only 8 percent of agencies had AI embedded in daily workflows, while 55 percent had no written AI policy.

The Big "I" and Liberty Mutual figures can look contradictory, but they measure different things. The Agency Universe count of 46 percent includes any AI use inside an agency, such as staff trying ChatGPT, while Liberty Mutual's 14 percent counts agencies that have formally implemented an AI tool.

Agencies cite a lack of knowledge as the main barrier. Sixty percent of agencies in the Agency Universe Study cited limited knowledge of what AI can do, and 48 percent pointed to security and privacy worries, according to Agency Checklists' summary of the findings.

Agencies want AI inside their existing systems

In a survey by the agency software firm AgencyBloc, covering health, life, Medicare and benefits agencies, respondents said they want AI built into the systems they already use. Only 24 percent had that; 83 percent were relying on general-purpose tools such as ChatGPT instead. Those respondents work in a different part of the market from the property and casualty agencies in the Liberty Mutual and Big "I" studies, and AgencyBloc sells agency software, so its figures are a pointer rather than a benchmark.

This is where standalone tools tend to fall short. An AI assistant can draft an email or summarize a policy, but if a producer has to copy client data out of the management system, paste it in, check the result and paste it back, the time saved shrinks and staff are less likely to keep using it.

Purdy's answer is a shift from a system of record, which simply stores information, to a system of work that notices a renewal is due, gathers what has changed and helps the broker decide what to do next. His company runs weekly focus groups with brokers and builds from their requests rather than dictating workflows, while acknowledging that some long-serving brokers will never embrace a shiny new tool.

Read next: Insurance futurist says AI is leveling, not disrupting, the field

Clients are adding pressure. Zywave's 2026 Broker Services Survey found that, for the first time, failing to adopt modern technology and AI made the list of top reasons employers give for switching brokers. Like AgencyBloc, Zywave sells AI software to brokers. Its survey nonetheless suggests some clients now judge agencies partly on their use of technology.

Questions to ask before buying

Based on the survey findings and Purdy's comments, principals considering a new AI product could put these questions to vendors and to their own staff:

• Does it work inside our agency management system, or does data have to be copied in and out?
• Which specific task does it handle, such as renewal preparation or certificate requests, and who does that work now?
• Where is our data stored, is it used to train the vendor's models, and how do we get it back if we leave?
• Which outputs need a human check before they reach a client, for errors and omissions purposes?
• Do we have a written AI policy covering approved tools and client data?
• Did the staff who will use it ask for it?

Read next: OpenAI sparks a distribution reckoning for insurance – what does it mean for brokers?

The surveys suggest agents want the time savings AI offers but have less interest in learning another standalone system. Vendors whose AI works inside the platforms agencies already use may find adoption easier. Purdy expects brokers to stay central either way. "The brokers are not going anywhere," he said.

With reporting by Branislav Urosevic