McKinsey's February 2026 report AI in Insurance: Understanding the Implications for Investors puts $50 billion to $70 billion of insurance revenue within reach of generative AI. The audience is portfolio managers and private equity partners, not chief underwriting officers, and the framing follows.
The number behind the headline is its scope. This is top-line across the whole value chain, brokers, MGAs, TPAs and vendors, not carrier expense savings.
Key Takeaways
- $50 billion to $70 billion is a revenue estimate across the full value chain, which is a different quantity from the $9.3 billion of P&C operating income uplift Morgan Stanley projects by 2030.
- Negative $2.4 billion is Morgan Stanley's net 2026 operating impact for P&C carriers, $3.0 billion of implementation cost against $600 million of realized savings, before the curve turns.
- 60% of EXL's Q1 2026 revenue came from data and AI-led work, growing 28% against 13.8% total company growth, up from 38% in 2020.
- $15 billion or more of insurance commissions are classified as low complexity and at risk from AI, per Bank of America, covering 20,000 to 30,000 independent agents.
- 20% of P&C insurers have a well-defined analytics strategy and 12% offer regular analytics training, per WTW.
Where McKinsey Puts the Money
The estimate is a sum over four subsectors with different economics, and the differences matter more than the total.
Brokers carry roughly 70% of private equity transaction volume in insurance, with the AI thesis resting on wallet share: automate submission processing, match carrier appetite algorithmically, serve more accounts per producer. MGAs are about 5% of deal activity but the steepest growth line, with US MGA direct premiums nearly doubling from $47 billion in 2020 to $97 billion in 2024, roughly 14% compounded, and AI compressing quoting from weeks to days.
Third-party administrators are the subsector where the thesis runs into its own commercial model. TPAs have grown around 15% annually on recurring revenue and deep servicing data, but pricing is generally tied to headcount or activity volume. A TPA that processes claims 40% faster earns 40% less per claim unless the contract moves to an outcome basis first. Software vendors, at roughly 20% annual investment growth, benefit from the shift toward modular architectures in which specialized agents coordinate rather than one system doing everything.
The private equity context is why the report reads as it does: US insurance-focused PE expanded at 26% annually between 2022 and 2025 while European invested capital fell 18% a year, and investors prioritizing operational value creation ran two to three points of IRR above peers.
Two Frameworks Pull the Expense Ratio in Opposite Directions
McKinsey and Morgan Stanley are not disagreeing. They are measuring different lines, and a pricing actuary has to know which one their carrier is executing before setting an expense load.
| Dimension | McKinsey (Feb 2026) | Morgan Stanley (Jan 2026) |
|---|---|---|
| Headline figure | $50B to $70B revenue potential | $9.3B operating income uplift by 2030 |
| Scope | Full value chain (brokers, MGAs, TPAs, vendors) | P&C carriers only |
| Metric | Revenue creation | Expense reduction |
| Audience | PE and institutional investors | Equity analysts covering public carriers |
| Near-term economics | Investment phase, elevated spend | J-curve: -$2.4B in 2026, positive by 2028 |
| Carrier expense ratio impact | May increase short-term (growth investment) | -200 bps by 2030 (from 30.5 to 28.5) |
The expense framework has a J-curve with a documented near term. Morgan Stanley's own numbers put P&C carriers at a net negative $2.4 billion in 2026, $3.0 billion of implementation cost against $600 million of realized savings, reaching a $9.3 billion annual uplift by 2030 through 200 basis points of expense ratio reduction. A carrier pursuing that path books higher expenses now for lower ones later.
The revenue framework inverts the near-term sign deliberately. A broker or MGA investing to serve more accounts per producer raises its expense base to buy premium growth, and the expense ratio improves through the denominator rather than the numerator. Applying an industry expense trend to a book following one strategy while the comparables follow the other produces an error in the same direction every year.
The vendor filings show the revenue side is at least measurable. EXL's data and AI-led revenue reached 60% of total in Q1 2026, growing 28% against 13.8% company growth on $570.4 million of revenue, up from 38% in 2020 and 55% for full-year 2025. CCC reported $281.3 million of revenue up 12%, with AI solutions at roughly $120 million annualized, about 10% of the total, growing at 3.5 times the company rate and margin-accretive at a 43% adjusted EBITDA margin. Verisk posted $783 million with 4.7% organic constant currency growth and seven new client-facing AI modules in the quarter.
Some of the Revenue Is Not New
The value-chain framing is what makes the estimate large, and it is also what makes part of it a transfer rather than a creation.
Bank of America's analysis puts $15 billion or more of insurance commissions in a low-complexity bucket exposed to AI, covering 20,000 to 30,000 independent agents whose work language-model agents could perform. The named commission pools are large: Progressive above $6 billion, Travelers around $3.35 billion, Hartford about $1.25 billion. Insurance distributor stocks had already fallen 24% from peak valuations, and BofA projected organic revenue growth slipping from a 3% to 7% band toward 1% to 5%.
Equity research has already started splitting the chain on that basis. Goldman Sachs upgraded AIG to Buy and downgraded Allstate to Neutral in March 2026, arguing that complex, multinational and large corporate risks require judgment that resists disintermediation while AI amplifies underwriting throughput on exactly those accounts.
Set that against the broker subsector carrying 70% of the report's transaction volume and the two views describe the same activity from opposite ends. Efficiency that lets one producer serve more accounts is the same efficiency that reduces the number of producers a book requires. Whether it nets to revenue creation depends on where in the chain the margin settles, which the aggregate figure does not resolve.
Capture also assumes an operational base that is not there yet. WTW's survey of P&C insurers found 20% with a well-defined analytics strategy and 12% offering regular analytics training, with data quality and IT support barriers each affecting 42% of respondents, against Sedgwick's finding that 7% of insurers have reached full AI scale. The estimate is a statement about an opportunity set. The distance between it and realized revenue is an execution problem most of the industry has not started on.
Further Reading on actuary.info
- McKinsey’s August 2026 Follow-Up: The Soft-Market Catch on Gen-AI’s Insurance Prize - The newer CEO-level framing of the same $50B-to-$70B estimate, set against 2026 rate data testing whether the modeled technical-result gain survives price competition.
- Morgan Stanley’s $9.3 Billion AI Savings Forecast for P&C Insurers - The expense-reduction counterpart to McKinsey’s revenue thesis, with carrier-by-carrier breakdown, J-curve economics, and 200-basis-point expense ratio projection through 2030.
- Deloitte, Oliver Wyman, and McKinsey Map Insurance AI Priorities for 2026 - Cross-framework synthesis showing where three consulting firms converge on AI as the top priority and where their combined ratio forecasts diverge by 2.1 points.
- EXL Q1 2026: AI Revenue Hits 60% and Reshapes the Insurance Vendor Model - The vendor-side revenue data that validates McKinsey’s thesis, with outcome-based pricing shifts and insurance segment detail.
- Carriers Build AI Expense Savings Into Forward Guidance for the First Time - How AIG, Chubb, Progressive, and Travelers are embedding AI savings into investor-facing projections, the operator complement to the consulting firm frameworks.
- Insurer AI Adoption Hits 82% But Only 7% Reach Full Scale - The execution gap between adoption and scale that constrains how quickly the McKinsey revenue estimate can be realized.