Insurance

AI for insurance brokers: renewals, submissions, and the service gap

Two-thirds of independent agencies plan to increase AI use this year. Only 8% have it embedded in daily work. That gap is the whole story, and the number one reason agencies get sued makes the case better than any vendor deck. For brokerages in Vancouver, Canada, and the US.

By Adi Huric, founder of Most AI Labs·July 2026·10 min read

Insurance brokerage is the rare industry where the trade association runs its own technology surveys, the errors-and-omissions claims data is published, and the regulator guidance is already written. That makes the AI story unusually easy to fact-check, and the facts say something specific: the profession has finished being curious and has barely started integrating.

Written for independent property and casualty brokerages and agencies, from two-person shops to mid-size firms, in Vancouver, anywhere in Canada, or the US. Every claim is cited. Vendor and carrier survey numbers are labeled as exactly that.

Adoption: the integration gap is the story

The Big I’s Agents Council for Technology surveyed member agencies in early 2026: 31% are not using AI at all, 33% are “just experimenting,” and only 8% have AI embedded in daily workflows. At the same time, two-thirds plan to increase AI use within twelve months. The tool mix explains the gap: 45% of agencies are using ChatGPT or other public chatbots, while purpose-built tools lag far behind (20% policy comparison, 13% document extraction, 13% chatbots).

The velocity is real, though. In Liberty Mutual’s 2024 study only 6% of agency principals had implemented any AI solution; its 2025 follow-up (1,200+ agency leaders and staff, a carrier-funded survey) found over a third already integrating AI into their work. From 6% to a third in roughly a year is fast. The head of the Big I’s technology council put the 2026 situation plainly: the gap is not technology, it is integration. Most agencies are not under-investing. They are under-integrating what they already have.

The E&O argument nobody markets

Here is the strongest business case in this article, and it comes from claims data, not a vendor. Swiss Re’s analysis of the Big I professional liability program found “coverage not procured” causes roughly 30% of agency E&O claims in both commercial and personal lines. Every other cause sits under 10%. The number one reason agencies get sued is a client who needed coverage, discussed it or should have discussed it, and did not end up with it. The program’s standing advice: you cannot over-document.

That is precisely what AI-assisted policy checking and documentation attacks. A renewal review that compares this year’s policy to last year’s, flags removed coverages, and logs the conversation is E&O defense wearing an efficiency costume. The same claims program reported a significant uptick in both frequency and severity of E&O claims through 2025, which makes the timing less than optional. The honest flip side: AI output that reaches a client unreviewed is itself an E&O exposure, which is why every regulator paragraph below insists on human review.

Where the work actually is

·Commercial submissions. 72% of agencies name the commercial submission process as the workflow they most want automated (2025 Ivans Connectivity Survey; Ivans is owned by Applied Systems, so treat it as vendor data measuring demand, not results). Re-keying ACORD forms between systems is the industry’s least defensible use of licensed staff.
·Renewal and policy review. The best documented example we found: an agency in the Big I’s 2025 Best Practices Study saving 30 minutes per staff member per day on AI policy checking. Our arithmetic, labeled as ours: that is roughly 130 hours per person per year, and about 1,300 hours across a ten-person agency, more than half a full-time position of recovered capacity.
·The service gap. Vertafore’s January 2026 policyholder survey (600+ US consumers, vendor survey): 83% expect a response within one business day, over a third expect one within an hour, and only 21% get proactive outreach from their agent. Only 13% of agencies run any chatbot or virtual assistant. The expectation-to-delivery gap is where books quietly leak at renewal.
·Cross-sell on the existing book is what agents themselves most want AI for (58% in the 2024 Liberty Mutual study). Interest data, not outcome data, but it points at the right asset: the book you already own.
·Certificates of insurance: we found no independent outcome data on COI automation at all, only vendor content marketing with untraceable benchmarks. Plausible workflow, unproven category. We would pilot it last.

The number one cause of agency E&O claims is coverage not procured. The claims program’s own advice is that you cannot over-document. AI that reviews renewals and documents everything is not a productivity toy. It is defense against the exact thing agencies get sued for.

The rules already exist. Most agencies have not written theirs.

In the US, the NAIC’s model bulletin on insurer AI use has been adopted by 25+ states as of late 2025. In Canada, the practical checklist for brokers is RIBO’s May 2025 guidance out of Ontario, the first broker-specific AI framework in the country: a human reviews AI output before it reaches a client, clients must know when they are talking to AI, client information never goes into open AI tools, and accountability stays with the licensed broker no matter what the software did. OSFI’s Guideline E-23 binds federally regulated carriers on model risk from May 2027, which will shape what carriers demand of AI-touched submissions.

BC brokers should know their own regulator is behind the curve: BCFSA has committed to draft AI guidance for consultation in 2026/27, and until then RIBO’s principles are the closest Canadian template. Meanwhile the client expectation is already set: 85% of clients want to know when their agent is using AI (Vertafore 2026). And the most fixable stat in this article: 55% of agencies have no written AI use policy. That document costs an afternoon and is the first thing we would produce with any brokerage client.

The talent squeeze behind all of this

The famous claim that half the insurance workforce retires by 2028 did not survive our fact-check (see the honesty note below). The defensible version is still sobering: recruiter analysis of US Bureau of Labor Statistics data puts roughly 1.37 million insurance professionals at age 55 or older against about 214,000 aged 20 to 24, better than a six-to-one ratio, and the Jacobson Group’s long-running labor study finds most insurance positions remain at least moderately difficult to fill in 2026. When you cannot easily hire licensed staff, recovering half an FTE from renewal reviews is not a luxury.

Watch out for this

Honesty note: this niche circulates several statistics we refused to use: “50% of the workforce retiring by 2028,” “400,000 unfilled positions,” and precise COI-automation savings figures. None of them trace to a primary source. If a vendor quotes them at you, that tells you something about the vendor.

The Canadian and BC picture

Canada’s broker channel is structurally strong: IBAC speaks for over 43,000 P&C brokers, the majority of Canadian P&C business is broker-distributed, and IBAC has stood up its own AI working group spanning insurers, brokers, and software vendors to produce broker guidance. In BC specifically, every vehicle owner in the province passes through a broker: ICBC Autoplan is sold through a network of roughly 900 Autoplan broker offices. Foot traffic and renewals other industries would kill for.

One more local detail we like: the renewal-automation tooling for this industry is being built in Vancouver. Quandri, headquartered here, raised US$12M in July 2025 with Intact’s venture arm on the cap table to automate policy and renewal reviews for brokerages (its customer results are vendor claims; the funding and investors are verifiable). The tools for this industry are local. The playbook can be too.

The honest bottom line

·Write your AI use policy first. 55% of agencies have none. RIBO’s four principles are a ready template: human review, transparency, no client data in open tools, accountability stays with you.
·Start with renewal and policy review. It has the best documented example (30 minutes per staff per day), and it doubles as E&O defense against the number one claim cause.
·Then submission intake. The demand is near-universal (72%) and the work is re-keying, the easiest thing to hand to software with a human check at the end.
·Fix response speed before adding chat. 83% of clients expect same-business-day answers. Meet that with routing and drafting assistance first; a chatbot without the workflow behind it is a complaint generator.
·Treat COI automation as unproven. No independent evidence yet. Pilot it last, on your own numbers.

If you want this mapped to your brokerage with your own book, renewal calendar, and E&O exposure, that is what our AI consulting work does. It starts with the free 7-day audit: where your renewals, submissions, and service actually leak, with the math shown before you spend anything. Fixed price. You own what gets built.

Wondering what an ad budget buys here? Our free lead calculator runs the verified benchmark math for this industry: insurance-category leads run about $74 on Google, and why we refuse to fake a quote-to-bind number.

Sources

If any of this is your week

Start with the 7-day audit.

7 business days. A real document. Yours to keep — whether you hire us or not.

Start with the 7-day audit Or talk to the founder first

AI for law firms: what the evidence actually supports

The same evidence-first treatment for another regulated, document-heavy profession.

Read→

Trades

AI for trades and home services: the missed-call goldmine

The verified missed-call math and speed-to-lead evidence.

Read→