McKinsey Puts Gen-AI's Insurance Prize at $50 to $70 Billion: The Soft-Market Catch

McKinsey estimates generative AI could unlock $50 billion to $70 billion of insurance industry revenue, concentrated in marketing and sales, customer operations, and software engineering, a figure it sets against two decades in which premiums compounded at 4.9% annually while pretax profit grew only 4.3% (McKinsey, August 2026). Whether that gain reaches the loss ratio or gets competed back to policyholders is a pricing question, not a technology one.

McKinsey's own survey work supplies the number that makes the pricing question unavoidable: more than 50 leaders from Europe's largest insurer groups told the firm gen AI could deliver 10% to 20% productivity gains and 1.5 to 3.0 points of technical-result improvement. Technical result is underwriting margin before investment income, the actuarial line item that separates a rate that is adequate from one that only looks adequate on a spreadsheet. If that improvement holds at the individual-carrier level, it is real margin. If every competitor captures the same 1.5 to 3.0 points at roughly the same time, in lines where rate filings are public and renewal shopping is frictionless, the improvement does not stay margin for long. It becomes the basis for next year's indicated rate decrease.

Where the $50 Billion to $70 Billion Concentrates

McKinsey's "How AI Will Reshape the Economics of Insurance" report, published in early August 2026, frames the opportunity against four structural problems it says have defined the industry since 2005: fading relevance, high distribution cost, flat productivity, and a slow pace of change. The relevance problem shows up as protection gaps that have widened even as the industry has digitized: a $133 billion natural catastrophe protection gap in 2025, and a cyber gap McKinsey puts at roughly $900 billion, since less than 1% of global cyber costs are currently insured (McKinsey, August 2026). The distribution problem is structural rather than cyclical. Roughly 85% of U.S. property and casualty premiums and 95% of life insurance premiums still move through agents, brokers, and managing general agents, channels that consume 10 to 25 cents of every P&C premium dollar and up to 80 cents of first-year life premium in commissions and acquisition cost (McKinsey, August 2026). The productivity problem is the one with the longest paper trail. McKinsey puts the average P&C expense ratio at a range of 27% to 32% since 2005, essentially flat for two decades, with a 10-point gap separating leaders running below 22% from laggards stuck at 32% (McKinsey, August 2026). That spread, not the industry average, is where the $50 billion to $70 billion estimate actually lives: McKinsey's own framing names marketing and sales, customer operations, and software engineering as the functions where gen AI closes the gap fastest, and those are exactly the cost centers that determine which side of the 22-to-32 divide a carrier sits on.

A Two-Decade Stalemate: Premiums Up, Profit Flat

The headline arithmetic behind the report is what gives the $50 billion to $70 billion figure its urgency. Global gross written premiums expanded at roughly 4.9% annually since 2005, reaching an estimated $8.3 trillion in 2025. Profits before tax grew at only about 4.3% over the same stretch, reaching approximately $580 billion (McKinsey, August 2026). A 60-basis-point gap compounded across 20 years is not a rounding error; it is the mathematical signature of an industry that has grown its top line without growing its margin, which is a different failure than shrinking. Reinsurance News summarized the report's blunt framing: the insurance industry of 2026 would remain recognizable to an executive who last saw it in 2006, a stagnation claim McKinsey backs with the observation that personal lines premium has actually fallen as a share of global GDP, from 1.2% in 2019 to 1% in 2023. McKinsey's answer is that AI does not just cut the expense ratio incrementally, the way prior technology waves did. It resets the two structural choices carriers have. The report describes two viable paths: scale, where large carriers amortize AI infrastructure across bigger premium bases, and deep specialization, where narrower players excel in a niche and buy the rest of the stack from partners. Carriers stuck between those two, running broad books without either scale or specialization, are the ones McKinsey implies absorb the 27-to-32% expense-ratio band indefinitely.

What Insurers Themselves Expect the Payoff to Look Like

The European survey figures give the abstract revenue estimate a concrete shape at the level of an individual carrier's income statement. Beyond the headline 10% to 20% productivity gain and 1.5-to-3.0-point technical-result improvement, respondents also pointed to premium growth of 1.5% to 3.0% attributable to AI-driven distribution and underwriting gains (McKinsey, August 2026). Those are self-reported expectations from executives with an obvious incentive to sound bullish to a consulting firm building an industry-wide business case, which is itself a reason to treat them as an upper bound on what any single carrier can expect, not a floor. Risk & Insurance reported the case study McKinsey uses to anchor those survey numbers in disclosed results: Aviva has deployed more than 80 AI models across its UK book, cutting liability assessment time by 23 days, improving claims routing accuracy by 30%, and reducing customer complaints by 65%, for roughly £60 million (about $82 million) in savings in 2024 alone. That is a single carrier's audited result, not a survey response, and it is the kind of figure that supports McKinsey's broader claim that AI leaders have generated 6.1 times the total shareholder return of AI laggards over the past five years, a performance spread wider than in most other industries it tracks (McKinsey, August 2026). The gap between Aviva's disclosed 2024 number and the survey's 1.5-to-3.0-point aspiration is the gap between a carrier that has already scaled AI past pilot and the much larger group that has not, a distinction actuary.info's coverage of scaled AI adopters found holds up in loss ratio data as well as in shareholder returns.

The Soft-Market Catch

McKinsey's technical-result estimate is a partial-equilibrium number. It models what one carrier gains from deploying AI while its competitors' behavior stays fixed. That is a reasonable simplification for sizing a single firm's business case. It is the wrong simplification for sizing an industry, because insurance pricing does not clear like a monopoly; it clears like a market where dozens of rated, regulated competitors quote the same risk and the buyer, retail or commercial, can shop the difference. When a cost or accuracy advantage becomes available to most of the market at roughly the same time, the standard result in a competitive market is that the advantage shows up in price, not in margin. The technical-result gain McKinsey's own survey respondents expect to keep is precisely the kind of gain a soft market is built to give away. 2026's actual P&C pricing data is the natural experiment. Commercial P&C premiums fell 1.2% in the first quarter of 2026, the first quarterly decline since 2017, according to the Council of Insurance Agents & Brokers market index, with commercial property down 5.5%, workers' compensation down 3.7%, and cyber liability down 3.5%. Brokers surveyed cited expanded underwriting capacity and more aggressive competition for both new and renewal business as the drivers, the textbook description of a market where carriers are using improved risk selection to win share on price rather than to defend margin. Personal auto tells the same story from the other side of the balance sheet: the net combined ratio improved to 94.4 in 2025, the strongest result in years, and premium growth simultaneously slowed to 3.6%, the slowest pace since 2020, according to Insurance Information Institute data. A carrier that keeps a technical-result improvement to itself grows share at stable rates. An industry that captures the same improvement broadly grows loss ratios more slowly and rate levels more slowly, at the same time, which is exactly what the personal auto numbers show. Cyber shows the mechanism running furthest along its life cycle. Global cyber rates fell for a 12th consecutive quarter through Q2 2026, down 4% year over year, even as ransom demands climbed 47% to more than $1 million on average, a pattern actuary.info's coverage of cyber rate adequacy traced to Marsh's placement data. Cyber underwriting has had the deepest AI-assisted analytics penetration of any commercial line for several years running, precisely the head start McKinsey's framework would predict produces the largest technical-result gain. Instead, filed rate has decoupled from loss-cost trend in the direction of price, not margin, because enough carriers reached similar underwriting sophistication at similar times that the gain competed itself away before it could compound. That is the mechanism McKinsey's CEO-level framing skips: a rate filing is not a business case memo. An actuary sets an indicated rate change off credibility-weighted loss experience and current trend, not off a projected productivity gain, and once enough of the market's loss experience reflects AI-assisted underwriting, the indicated change for everyone converges lower, whether or not any individual carrier wanted to give the margin back.

Build, Buy, or Fall Behind

The 6.1-times TSR gap between AI leaders and laggards is, in that light, less a permanent structural advantage than a measure of how long a first-mover window stays open before competitive pricing closes it. Two data points on the vendor side of the industry sharpen that read. BCG's 2026 AI Radar found P&C carriers have tripled AI spending as a share of revenue, yet only 38% of them are generating measurable value at scale, a gap actuary.info covered in detail when the survey was released. Morgan Stanley's competing bottom-up forecast projects a 200-basis-point P&C expense ratio reduction and $9.3 billion in industry operating income by 2030, but only after a net negative $2.4 billion impact in 2026 as implementation costs outrun realized savings, a J-curve actuary.info stress-tested against carrier-level assumptions. Both figures point the same direction: the margin McKinsey's survey respondents expect to keep is not guaranteed by deploying AI. It is guaranteed, if at all, by deploying it faster and more completely than the 62% of carriers BCG found are still stuck below scale. That gap has a direct build-versus-buy consequence for actuarial and pricing functions specifically. A carrier that licenses the same off-the-shelf gen-AI underwriting or claims-triage tool its competitors are also licensing gets to the 10-to-20% productivity gain fastest, but it also commoditizes fastest, because every licensee of the same vendor model is generating the same technical-result improvement on roughly the same timeline, the precise condition under which a soft market converts that improvement to price rather than margin. A carrier that builds proprietary capability, harder and slower, retains a differentiated technical result for longer, because its competitors cannot buy the same edge off a shelf. Pricing actuaries setting expense loads and rate indications for the next several renewal cycles need to know which path their carrier is on, since a rate plan built on the assumption of a durable AI-driven technical-result advantage is making a bet about vendor commoditization that the 2026 cyber and personal auto data suggests does not hold for long once the tooling becomes widely available.