Blog
Will AI replace insurance brokers? A straight answer
AI will take over parts of a broker's (US: agent's) work, mainly the reading, copying and comparing. Nobody can say how fast, and anyone who gives you a date is guessing. The advice, the client relationship and the accountability still rest with a person, and the rules say so. The sensible move is to prepare, calmly.
Key points
- AI will take over parts of a broker's work, mainly the reading, copying and comparing.
- Nobody can say how fast, and anyone who gives you a date is guessing.
- The advice, the client relationship and the accountability still rest with a person.
- Prepare by learning the tools, moving one job to software with a person checking, writing the data rules and keeping time for clients.
Will AI replace insurance brokers?
AI is unlikely to replace the broker as a whole, but it will change what a broker spends the day on. We can't prove that, because it is a forecast, and forecasts about technology are often wrong. What we can say is which tasks software can already do, and who is still answerable when something goes wrong.
A signed Insurance Times briefing by Mike Edgeley, group chief executive of Clear Group, published on 25 February 2026, argues the same way about commercial brokers: "It will not eliminate the broker, but it will redefine what great broking looks like." He runs a broking group, so he has a stake in that view. Read it as one informed opinion, not a measurement.
What can AI already take over?
AI can already take over the preparation work: reading documents, pulling out details, comparing two versions and drafting a note. "AI" here means software that reads and writes text. The table shows the tasks where that matters most to a broker, and where a person stays involved.
| Task | What AI can take over | What stays with a person |
|---|---|---|
| Renewals | Gathering changes, comparing wording with last year, listing gaps | Advising the client on what to do |
| Quotes | Lining up premium, excess (US: deductible), limits and exclusions | Recommending which cover to buy |
| Policy checks | Reading the issued policy against the agreed terms | Deciding what to do about a difference |
| Policy changes | Turning a request into a complete change form | Confirming facts with the client |
| Claims | Keeping one record per claim, listing what is outstanding | Speaking up for the client with the insurer |
| Finance | Matching commission lines to your records | Resolving disputes with an insurer |
Each of these has its own guide, such as preparing renewals and comparing quotes. In every case the same pattern holds: untidy information arrives, software tidies it, a person checks it, and a person decides.
The honest part is that some of that work is what junior staff learn on. If software does the reading, an office has to find a new way to train people. That is a real question for owners, and nobody has a settled answer yet.
What do clients and regulators still expect from a person?
Clients and regulators still expect a named person to give the advice and answer for it. The FCA, the UK regulator, says on its page on its approach to AI (first published 8 September 2025): "We do not plan to introduce extra regulations for AI. Instead, we'll rely on existing frameworks, which mitigate many of the risks associated with AI." In plain terms (our reading), the existing rules apply to your work whether a person or a program prepared it.
The same page lists the Consumer Duty, which says UK firms must "Design products and services that meet the needs of their target customers and provide fair value." Software cannot hold that duty. Your firm does.
In the US, insurance is regulated state by state. The National Association of Insurance Commissioners (NAIC) publishes a model bulletin on insurers' use of AI, and its AI page says: "Human oversight remains an important part of insurance decision-making." The NAIC adopted its model bulletin on 4 December 2023. It is addressed to insurers, so it does not bind a broker directly, and it asks insurers to cover third-party AI systems. Your state may have adopted it in its own form, and it may reach you through the insurers you work with, so ask your state regulator or your compliance adviser. This page is not legal advice.
On jobs, the US Bureau of Labor Statistics projects employment of insurance sales agents to grow 3 percent from 2025 to 2035, about as fast as the average for all occupations, from 572,600 jobs in 2025 (page last modified 27 August 2026, read 5 November 2026). A projection is an estimate, and it covers US sales agents only.
Clients matter here too. A client who has had a claim declined wants a person who knows their business, speaks up for them and picks up the phone. That is the part of broking that software does not carry, and it is the part clients pay for.
What stays with a licensed person?
Six things should stay with a named person who holds the licence or authorisation your market requires.
- Advising the client.
- Recommending which cover to buy.
- Confirming facts with the client.
- Deciding what to do about a difference between a policy and what was agreed.
- Answering to the client.
- Answering to the regulator.
This is good practice, not a universal law. It fits the FCA and NAIC pages quoted above: existing rules still apply, and human oversight remains important. A named person should also stay accountable for any quote given or cover placed, whatever software prepared the paperwork.
What should a broker do about it?
A broker should do four modest things now: learn what the tools do, move one repeating job to software with a person checking, write down the data rules, and keep time for client conversations. None needs a big project. Each is reversible if it does not work.
- Learn by using. Try a general assistant on everyday writing with no client data. Our guide on using AI in an insurance agency lists safe first uses.
- Pick one job. Choose the repeating task that costs your office the most hours, and measure it. The 90-day AI plan shows how.
- Write the data rules. Decide what may go into an AI tool before anyone does it. See whether ChatGPT is safe for client data.
- Protect the human part. Use the hours you get back for the conversations that win and keep business.
It also helps to be sceptical in both directions. Be wary of anyone who says brokers are finished, and equally of anyone who says nothing will change. Judge claims by the task they describe, the evidence offered and whether the person making them has something to sell. That includes us.
The main guide to AI for insurance brokers gives the wider picture, and the guide to AI tools for insurance agents helps you choose between tools, a built system and hiring.
Where Navicade fits
Navicade builds and runs AI systems that do work the office does by hand, so your people keep the advice and the clients. We start with a free 30-minute call and a free operations map of where your hours go. Then we take on one agreed task at a time. Results arrive in your existing tools. Insurance decisions stay with your team. The system checks its work on every run and alerts us.
A beginner may not need us yet. If you have not tried a general assistant or measured a job, start there. When a repeating job is the problem, ask for the free call and map.
Next step
Start with a free map of your office.
Navicade builds and runs AI systems for jobs like this one. It starts with a free 30-minute call and a free operations map.