Insurance AI Pivots From Claims Efficiency to Underwriting: What the ILTF 2026 Shift Means for Carriers

Datos Insights surveyed 36 senior carrier technology leaders ahead of its April 2026 Insurance Leaders Technology Forum in Boston. The share with AI in production rose from 37% to 61% in a year, and for the first time underwriting deployment (56%) sits ahead of claims (50%).

The six-point gap is small but it is a reversal. Through 2024 claims led underwriting in every comparable survey, because claims workflows were the ones narrow models could handle.

Key Takeaways

• Production AI went from 37% to 61% of surveyed carriers in twelve months, a 24-percentage-point move that ends the pilot phase rather than extending it.
• Underwriting AI is live at 56% of carriers against 50% for claims, the first survey in which the order is reversed, and four of the top five deployed use cases are still document processing.
• Only 8% of respondents believe they lead peers on AI capability, while 70% expect moderate competitive advantage within three years.
• 70% of carriers spend under $500,000 a year on AI, against $250,000-$350,000 for a single senior data scientist in New York, so most production deployments are vendor-supplied by budget necessity.
• Consumer support for AI in P&C nearly doubled to 39%, but only 22% accept AI filing a claim for them and 16% accept AI cancelling or renewing a policy.

What the ILTF Survey Actually Shows

The pre-conference survey of 36 senior technology leaders is small but pointed. Production AI moved from 37% to 61% of respondents in a single year. Carriers are no longer evaluating.

Maturity is a different question. Only 8% believe they currently lead their peers, while 70% expect AI to deliver moderate competitive advantage inside three years. That is a large population betting on returns it has not yet measured.

The use case mix says why. Four of the top five deployed use cases are document processing: reading, extracting, summarizing, classifying. That is table-stakes automation rather than differentiation, and it is where the 61% figure mostly lives.

The function split is the finding that matters. Underwriting AI is live at 56% of surveyed carriers against 50% for claims. Six points is not a gulf, but the ordering has flipped, and it flipped in the same quarter that Evan Greenberg described "agentics within AI" as a five-year growth catalyst on Chubb's Q1 call and Peter Zaffino detailed an orchestration layer coordinating multiple agents across AIG.

Why the Loss Ratio, Not the Expense Ratio, Moved the Money

Claims automation was first because its wins were measurable: faster FNOL intake, automated damage estimation, straight-through processing on low-complexity claims. After three to four years those wins are captured, and each new project buys less.

Underwriting pays differently. It moves the loss ratio at the point of risk acceptance rather than the expense ratio after a loss has occurred.

Metric	Pre-AI Baseline	AI-Enabled (2026)	Improvement
Straight-through processing rate	10-15%	70-90%	5-7x increase
Quote-to-bind cycle time	3 days (typical)	Minutes	Up to 99% reduction
Loss ratio impact	Baseline	3-5 point improvement	Direct underwriting income
P&C underwriting expense ratio	Baseline	15-20% decline projected	Structural cost reduction
Life underwriting expense ratio	Baseline	25%+ decline projected	Larger automation scope

The arithmetic is what redirected the budgets. Commercial P&C insurers running agentic underwriting report loss ratio improvements of 3-5 points. A 4-point improvement on a $500 million premium book is $20 million of annual underwriting income, against a marginal claims automation project measured in processing minutes. AIG's Lexington unit processed more than 370,000 submissions in 2025 on generative AI with a target of 500,000 by 2030, and AIG Assist has reported a 55% time-to-quote reduction and a 40% binding lift across eight commercial lines.

That improvement is not evenly available, which is the part that should concern a pricing actuary. If one carrier's selection is materially better than its competitors', it wins the better risks and leaves the residual to everyone else. The winner's curse is an old dynamic; what is new is the speed at which a 70-90% STP book can execute it. Chubb is deliberately incremental here, embedding AI into an operating model already producing an 84% combined ratio, while Hartford grew Business Insurance written premium 6% at an 89.2% underlying combined ratio with an AI assistant inside the underwriting workflow.

Production Scale on Experimentation Budgets

The constraint sits in the spending line. With 70% of carriers under $500,000 a year and a single senior data scientist costing $250,000-$350,000, most of the 61% production figure is vendor-supplied capability rather than built capability.

That turns an operating metric into a counterparty exposure. A carrier running 70-90% STP through a third-party platform has tied its expense ratio and its processing capacity to a vendor's continuity. Kurt Diederich, CEO of Finys, drew the parallel in Carrier Management the week after the forum: the current landscape of high entrant volume and uneven differentiation resembles the early 2000s internet proliferation, where only a small share of vendors proved durable.

For an actuary, that belongs in the enterprise risk assessment rather than the technology roadmap. The question is what happens to the expense assumption and the quote capacity if the platform is acquired, repriced, or terminated, and whether the pricing model's documented assumptions identify which AI components are load-bearing.

The other limit is on the demand side. An April 2026 Insurity survey of over 1,000 US adults found P&C consumer support for AI nearly doubled from 20% to 39% in a year. Support falls away as the decision gets more consequential: 22% are comfortable with AI filing a claim on their behalf, 16% with AI cancelling or renewing a policy. Commercial and specialty buyers evaluate AI capability as a purchasing criterion; personal lines policyholders do not, which is why the most aggressive underwriting deployments cluster where they do.