The Greatest Threat the Independent Agency Has Ever Faced: AI disintermediation estimates are almost certainly too low, and underestimated commission losses, carrier self-interest, and the speed of large language model deployment are conspiring to imperil the independent agency model of distribution
The April 13, 2026, editor’s note in Insurance Journal frames a now-familiar reassurance. Yes, artificial intelligence will disintermediate some independent agency business, but the exposed slice is small, the complex middle market remains safe, and the independent agent will retain her place at the center of insurance distribution (Wells 2026). That reassurance rests on a Bank of America Global Research estimate that roughly $15 billion of the more than $100 billion in U.S. independent agency commissions is at near-term disintermediation risk, about fifteen percent of the pool, concentrated in low-complexity personal lines and small commercial accounts (BofA Global Research 2026; Fortune 2026). To put this assessment and the suggestion that the “independent agent will retain her place” in proper perspective, allow me to say that it is all bullshit. The fifteen-percent figure is a floor, not a ceiling. The reports systematically understate both the share of commissions that will migrate away from human intermediaries and the willingness of carriers to assist in that migration. Taken together, generative AI and collaborative carriers represent the most serious structural threat to the independent agency model of distribution that has ever appeared on the horizon. The fact of the matter is that most intermediaries will need to find new professions.
Why the BofA Estimate Understates the Commission Pool at Risk
The headline number, $15 billion, is derived from a sample of only six carriers (Travelers Select and Personal, Hartford Small Business and Personal, Progressive, Cincinnati Financial, Hanover, and Selective) and is explicitly limited to business that the analysts could classify as ‘low complexity’ from public disclosures (BofA Global Research 2026). The analysts themselves concede the limitation. ‘Public insurance agent/broker disclosure tends to be very thin, making it difficult for investors to estimate what proportion of their commissions skew toward low complexity,’ and they single out Liberty Mutual as a carrier generating ‘additional billions in commissions that also can skew toward low complexity’ but excluded for lack of data (BofA Global Research 2026). A floor estimate built from a six-carrier sample with thin disclosure is not a ceiling on industry exposure; it is a measurement artifact.
Three structural realities push the true figure substantially higher. First, the consolidation of the brokerage industry through aggressive ‘tuck-in’ acquisitions has, in BofA’s own words, produced a ‘snowball effect’ in which hundreds of acquired retail shops have quietly moved low-complexity, small-ticket business onto the books of the largest publicly traded brokers, where it is invisible in the disclosed line-of-business mix (Fortune 2026). Second, the line between ‘simple’ and ‘complex’ commercial accounts is not a wall but a gradient. A 25-employee professional services firm with a BOP, workers’ compensation, commercial auto, and a modest umbrella looks complex to a producer’s commission statement and looks routine to a sufficiently capable language model with API access to a rating engine. Third, even where an AI agent does not write the policy itself, the Bank of America analysts note a separate and underappreciated channel of value erosion: AI will ‘demystify insurance markets’ for sophisticated buyers, producing a ‘deflationary impact’ on broker pricing power even where the broker remains nominally in the transaction (BofA Global Research 2026). Commission compression is disintermediation by another name.
Why Carriers Will Cooperate in the Demise of the Independent Agent
The most consequential omission in the reassurance literature is its assumption that carriers are passive bystanders—or, better, loyal allies—of the independent agency channel. This assumption flatters the relationship and misreads the economics. Commissions are the single largest controllable acquisition expense on a property-casualty income statement. In personal lines, agency commissions typically run between ten and fifteen percent of premium; in small commercial, they often run higher. A carrier that can substitute a marginal-cost API call to a large language model for a fifteen-percent commission has discovered, in effect, a permanent rate increase it does not have to file.
The behavioral signal is already visible. In February 2026, Spanish digital insurer Tuio became the first regulated carrier to deliver real-time bindable homeowners quotes inside ChatGPT, and the U.S. comparison platform Insurify launched a parallel motor-insurance experience inside the same chat interface (Leader’s Edge 2026; Genasys Technology Group 2026). The market reaction was instructive: on February 9, public broker stocks fell roughly nine percent in a single session, with Willis Towers Watson down approximately 13 percent, Gallagher 9.4 percent, and Aon 8.5 percent (Leader’s Edge 2026). The stocks have only partially recovered. Equity markets do not punish a sector to that magnitude based on a single foreign homeowners app unless they suspect that something larger is being signaled, namely, that the carrier-to-AI distribution stack is now technically and commercially viable, and that the cost structure it implies is incompatible with paying a human intermediary fifteen percent in perpetuity.
The public posture of carriers, dutifully thanking ‘our independent agent partners’ at every industry conference, should be understood for what it is. A sort of bullshitty bridge strategy. Carriers will continue to praise the channel for as long as the channel produces the bulk of premium, and they will quietly build the infrastructure that allows them to bypass it the moment a credible alternative exists. The history of distribution disintermediation in adjacent industries is unambiguous. Travel agents were assured for two decades that airlines valued their relationships. Bank of America itself draws the analogy to print advertising, noting that ‘Facebook/Meta and Google/Alphabet did not replace print advertising overnight but that over 20 years, consumer behavior changed to dramatically shrink the print ads market’ (Fortune 2026). Independent agents should expect a similar arc, compressed into a shorter window because the deployment cost of a language model is, as BofA observes, ‘cheap, easy, and happening right now’, a stark contrast to the trillion-dollar infrastructure cost of, say, autonomous vehicles (Fortune 2026).
Why This Threat Is Categorically Different from Earlier Disruptions
Independent agencies have weathered direct writers, captive networks, online comparison sites, and the first generation of insurtech. Each was survivable for a recognizable reason. Each required either heavy capital, a narrow product focus, or a customer willing to abandon the convenience of a single human point of contact. Generative AI eliminates all three constraints simultaneously. A large language model with tool-use access to carrier APIs can quote across hundreds of carriers, explain coverage in natural language, walk a buyer through endorsements, and deliver a bindable quote in the same conversational register that an agent provides (at a marginal cost approaching zero and at any hour) (Genasys Technology Group 2026). Gartner has forecast that traditional search-engine query volume will fall by approximately twenty-five percent by 2026 as consumers shift to AI assistants. ChatGPT alone surpassed 5.8 billion monthly visits by mid-2025 (Metricus 2026). The starting point of the insurance-buying journey is migrating into a channel that the independent agent does not control, does not appear in by default, and cannot easily buy visibility within.
The visibility problem deserves emphasis because it is the mechanism by which the BofA estimate will be exceeded faster than its authors expect. Independent agents have historically captured business through local search, referrals, and incumbency. When a consumer asks an AI assistant ‘what is the best homeowners policy for a Florida coastal property,’ the assistant returns a narrative recommendation naming specific direct carriers; it does not, by default, surface the local independent agency three miles away (Metricus 2026). At an industry-average commission of $300 to $500 per personal lines policy, a typical agency placing 500 to 1,500 personal lines policies stands to lose between $7,500 and $37,500 in annual commission revenue for every five percent of local consumers whose buying journey now begins with an AI assistant rather than a search engine (Metricus 2026). That is not a tail risk; that is the base case for 2027.
What This Means for the Independent Agency Model
The reassurance literature offers independent agents a familiar prescription, i.e., move upmarket, become a ‘trusted adviser,’ specialize in complex commercial risk (see my blog about this one), and treat AI as a productivity tool rather than a competitor (PwC 2025; Genasys Technology Group 2026). This advice is not wrong, but it is insufficient as a defense of the independent agency model as a whole. There are roughly 36,000 independent agencies in the United States, and the great majority of them derive the great majority of their revenue from precisely the personal lines and small commercial business that AI is most capable of absorbing. A strategy that preserves the top quintile of agencies—those with genuine specialty expertise and a complex book—while ceding the bottom four quintiles to direct AI distribution is not a defense of the channel. It is a managed liquidation of it.
My honest framing is this. The independent agency model has survived every prior disruption because no prior disruption simultaneously matched human intermediaries on cost, on availability, on multi-carrier breadth, and on conversational explanation. Generative AI matches all four, and it does so with a deployment cost measured in API tokens rather than infrastructure. Carriers have every economic incentive to assist the transition, and the public equity market has already begun to price it. The $15 billion figure that anchors the current debate is a conservative measurement of a sample, not a forecast of an outcome. The actual at-risk pool, once mid-market commercial deflation, carrier-led direct strategies, and AI-driven shifts in consumer search behavior are included, is materially larger, and the timeline over which it will be realized is materially shorter than the industry’s reassurance literature suggests. And this ALL rests on some dubious terminology about “low complexity”. Who determines THAT? What may have required professional tradecraft, “judgment underwriting” in our parlance, will most certainly fall victim to “mission creep”, A.I. itself determining independently that a complex submission really isn’t so complex after all. Independent agents who plan around the floor estimate will discover, as travel agents did before them, that the floor was the ceiling of an optimism they could not afford.
~ C. Constantin Poindexter, MA, JD, CPCU, AFSB, ASLI, ARe, AINS, AIS, CPLP
References
- BofA Global Research. 2026. Putting Numbers Around Insurance Agent/Broker AI Disintermediation Risk. New York: Bank of America Securities.
- Fortune. 2026. “$15 Billion of the Insurance Industry Is at Risk from AI, BofA Says.” March 3. https://fortune.com/2026/03/03/15-billion-of-the-insurance-industry-is-at-risk-from-ai-bofa-says/.
- Genasys Technology Group. 2026. “AI Insurance Distribution: How Brokers Will Win in 2026.” March. https://www.genasystech.com/ai-insurance-distribution-brokers-2026/.
- Leader’s Edge Magazine. 2026. “Insurance Brokers Plunge on AI Fears.” March. https://www.leadersedge.com/industry/insurance-brokers-plunge-on-ai-fears.
- Metricus. 2026. “Why Doesn’t AI Recommend My Insurance Company? Carrier Visibility Data.” April. https://metricusapp.com/blog/insurance-ai-visibility/.
- PwC. 2025. Insurance 2030: Next in Insurance. New York: PricewaterhouseCoopers.
- Wells, Andrea. 2026. “AI Disintermediation.” Insurance Journal, Southeast Edition, April 13, p. 6. https://www.insurancejournal.com/magazines/mag-editorsnote/2026/04/13/865249.htm.