States Are Drawing Lines on AI Claims Decisions. The Map Is Getting Complicated.
Three states have now enacted laws that directly regulate how health insurers can use artificial intelligence in claims coverage decisions, and the frameworks they have chosen are meaningfully different from each other. A Mondaq analysis published this week from attorneys at Taft Stettinius & Hollister lays out the current state-level landscape clearly.
Nebraska’s law prohibits utilization review agents from basing coverage decisions solely on an AI algorithm and requires separate disclosures to the Department of Insurance, health providers, and enrollees. Colorado’s statute takes a broader approach — applying to “any insurance practice” and prohibiting any algorithm or predictive model that produces unfair discrimination based on race, national origin, disability, or other protected characteristics, with officer-level attestation of compliance required. Georgia threads the needle differently: it permits AI as part of utilization review, but expressly prohibits AI systems from issuing an adverse determination until a qualified human clinician conducts a review. The statute’s language is direct — AI “shall not supersede the judgment of such clinical peer.”
What the three frameworks share is the assumption that fully automated adverse decisions are not acceptable in health insurance claims, regardless of model accuracy. That is a policy choice with real underwriting and operational consequences. Carriers running AI-assisted prior authorization workflows — and there are many — now have to navigate three distinct compliance regimes across three states, with more likely to follow.
Why This Matters for Insurance:
The Georgia model is worth watching closely. It effectively mandates human-in-the-loop architecture for any AI system that touches adverse claims outcomes. That is not a governance preference — it is a statutory requirement. Carriers that have built workflow automation around AI-generated recommendations without embedding a clinical review gate face genuine compliance exposure in Georgia, and potentially in the next state to pass a similar law. For the claims technology vendors serving health insurers, these statutes are shaping product requirements in real time.
Risk & Insurance: The D&O and E&O Market Is Scrambling to Define What It’s Actually Covering.
A detailed piece published this week by Risk & Insurance captures what is happening inside the specialty insurance market as AI risk moves from theoretical to actual. The core problem is definitional: carriers cannot underwrite AI risk if they cannot define what AI is, and the industry has not reached consensus.
Munich Re and HSB have chosen a deliberately broad definition — any statistical model that produces an output or takes autonomous action — specifically to avoid the silent-AI problem that haunted cyber insurance in its early years. The concern is legitimate. If traditional policies do not explicitly mention AI, courts may interpret that silence as coverage, creating unpriced liability the carrier never intended to take on. MSIG’s Bryan Boyle noted that the industry expects AI coverage to rationalize across product lines much the way cyber did — starting with ambiguity and coordination problems, eventually settling into cleaner standalone or endorsed structures.
The underwriting process is changing too. Westfield Specialty’s Jeff Kulikowski described underwriters now looking at the propensity of underlying models — including public models like Gemini and ChatGPT — for fraudulent outputs or discriminatory results. Ray Ash, EVP of financial lines at Westfield Specialty, put the board-level obligation plainly: companies now have to disclose AI exposures to investors the way they disclose cybersecurity risk. The 28% of public companies that have not yet disclosed AI risk in their 10-K filings are sitting on an unsecured exposure. The article draws a direct comparison to class actions that followed major cyber breaches a decade ago.
AI washing is emerging as a D&O trigger. Carriers are seeing companies promote AI capabilities that do not exist, and if those claims drive investor decisions and fail to materialize, they meet the standard for securities class actions.
Why This Matters for Insurance:
The specialty market is actively constructing the coverage architecture that will govern AI liability for years to come. Carriers that move carefully — building definitions that capture the full scope of machine learning without opening gaps — are in a better position than those waiting for consensus to emerge. The market’s current state closely resembles cyber insurance circa 2012: definitionally unsettled, increasingly exposed, and beginning to price seriously for the first time.
AllDigital Specialty Runs 70% of Its Business on AI. Here Is What Established Carriers Can Learn From How It Got There.
Insurance Business America published a detailed interview this week with Athula Alwis, CEO of AllDigital Specialty Insurance, a company that started AI-first and has never had to run a transformation project. That structural advantage is worth understanding.
AllDigital handles 70% of its volume autonomously: roughly 30 to 40% receives an automated approval, another 30% is declined outright by AI systems, and only the remaining 30% goes to human review. Alwis frames this through a four-stage AI maturity model: recommendation, assistance, execution, and orchestration (agent-to-agent). The company operates at stage three. Two of AllDigital’s six co-founders are AI specialists, and the company holds a U.S. patent for training machine learning systems in the specialty insurance sector.
Alwis is direct about what larger carriers get wrong. About ten years ago, companies brought in expensive data scientists who did not understand loss ratios. It took years, cost a lot, and mostly failed. His prescription for established carriers is a hybrid team: internal staff who understand legacy data and terminology, paired with engineers who know how to build modern machine learning systems. And he identifies data silos as the single biggest cultural barrier — citing the simple example of two underwriters calling the same expense item “legal expenses” and “ALAE” in different systems, producing inconsistent model inputs.
Human governance has not been reduced despite the high automation level. “We need human governance in everything we do,” Alwis said. “Model governance, changes, and guardrails are all monitored by human experts.” He treats that oversight not as a regulatory concession but as a design principle.
Why This Matters for Insurance:
AllDigital Specialty is a benchmark, not a template. Most established carriers cannot replicate a greenfield build. But Alwis’s diagnosis of where large-carrier AI initiatives fail — insufficient domain expertise, data silos, talent mismatches — is consistent with what practitioners across the industry are reporting. The 80,000-engineer AI talent shortage he cites makes external hiring alone an unreliable strategy. Carriers that invest in growing AI expertise internally, even on a slower timeline, are building something that does not leave when a recruiter calls.
Venture Capital Just Put $46 Million Behind AI-Driven Insurance Back Office. The Investor Roster Is Worth Noting.
Pace, an AI operations platform built specifically for insurers, announced a $46 million Series B this week co-led by Thrive Capital and Sequoia Capital, with participation from Emergence Capital and Pruven Capital. The company has already processed more than 250,000 insurance workflows using AI agents and counts Prudential Financial, WTW, and Newfront among its partners.
The Pace model is worth understanding in some detail. Its AI agents navigate internal applications, reason across documents, and — importantly — make phone calls to complete back-office insurance tasks including submission intake, policy servicing, claims handling, and data entry. At Prudential, Pace is automating thousands of hours of manual work tied to policy servicing. In partnership with Ryze Claim Solutions, it has reduced claim cycle times by 30%. At Convex US, it accelerates data ingestion for new business and renewals.
The investor roster matters as much as the dollar amount. Thrive Capital is closely associated with OpenAI-adjacent investments. Sequoia Capital is the defining venture institution for enterprise software bets. When both firms co-lead a Series B in an insurance AI company, it reflects a specific thesis: that insurance back-office operations represent a large, durable automation target that the general AI wave has not yet reached. Emergence Capital, a specialist in enterprise cloud software, brings a different kind of signal — this is a category they understand well, and they are betting it looks like other enterprise software markets that were transformed by AI agents.
Why This Matters for Insurance:
The protection gap that Pace’s CEO cites — $9 trillion globally — will not close through better front-end product design alone. It requires operational capacity: the ability to process more submissions, service more policies, and resolve more claims without proportional headcount growth. That is the problem Pace is built to solve. Carriers evaluating AI operations platforms should watch how Pace performs against its Prudential and Convex commitments over the next 12 to 18 months. The Series B gives the company enough runway to prove the thesis at scale.
AllianceBernstein: AI Is Repricing the Liabilities Side of the Insurer Balance Sheet, Not Just the Operations Side.
Most AI-in-insurance coverage focuses on operational efficiency — claims automation, submission processing, underwriting speed. A new AllianceBernstein research piece published this week argues that the more consequential impact on insurers is structural, cutting across multiple lines simultaneously.
The AB authors, drawing from the firm’s April 2026 Rethinking Insurance Forum, identify four specific exposure areas where AI is changing the underlying risk model: property and casualty books are being repriced by climate volatility accelerated by AI-driven data processing; autonomous vehicles are beginning to alter motor insurance economics; AI-enabled medical advances are affecting longevity assumptions in life and annuities; and cyber risk is scaling as AI strengthens both attackers and defenders in parallel.
The concern AB raises is one that actuaries are already grappling with: traditional siloed approaches to underwriting, reserving, and portfolio construction may be inadequate when AI-driven changes affect multiple lines at once, in correlated ways. An AI-enabled medical advance that extends average life expectancy by two years affects life insurance, long-term care, annuities, and health in connected ways that separate reserve calculations do not easily capture.
On the investment side, AB’s analysis suggests the macro backdrop is becoming less favorable for traditional fixed-income-heavy insurer portfolios, and makes a case for higher allocations to real assets and private credit as an inflation hedge. That is a portfolio construction argument, not an AI-specific one — but it is relevant context for CFOs and CIOs managing the asset side of an insurance balance sheet under sustained uncertainty.
Why This Matters for Insurance:
The AB piece is the kind of analysis that tends to reach the C-suite of major carriers precisely because it comes from a respected institutional asset manager rather than a technology vendor. The framing — that AI is repricing the liability side of the balance sheet, not just reducing the cost of operations — is one that chief actuaries and CROs should be tracking. The interaction effects across lines are likely to matter more than any single AI-driven development in isolation.
From the AI World: Microsoft’s AI Chief Sets an 18-Month Clock on White-Collar Automation
Mustafa Suleyman, CEO of Microsoft AI, told the Financial Times that most tasks involving sitting at a computer — accounting, legal, marketing, project management — will be fully automated by AI within 18 months. He named exponential growth in computational power as the primary driver, and described his core mission as achieving AI “superintelligence” and reducing Microsoft’s reliance on OpenAI in favor of proprietary frontier models.
Suleyman’s timeline should be treated with the skepticism that all such predictions deserve. A few months after he made that statement, Fortune noted, “mounting evidence shows AI is kind of a bust” in the broad economy — a reference to Apollo chief economist Torsten Slok’s finding that while Big Tech profit margins rose more than 20% in Q4 2025, the broader Bloomberg 500 has seen almost no change. A nonprofit study of software developers found AI actually made their tasks take 20% longer in at least one controlled setting. And nearly 80% of white-collar workers, per Fortune’s reporting, are outright refusing AI adoption mandates.
None of this means Suleyman is wrong about the trajectory. It means the pace of disruption is harder to predict than the direction. Insurance is not a sector that should be betting on an 18-month timeline — but neither should it be betting that disruption is a decade away.
Why This Matters for Insurance:
Insurance employs a lot of people doing exactly the tasks Suleyman describes: documentation, compliance review, policy processing, underwriting support, and claims administration. The ambiguity in the timeline is not an excuse to defer decisions. What it argues for is the same posture that applies to any strategic uncertainty — building organizational capacity to adapt as evidence accumulates, rather than waiting for certainty that will not arrive before the decision is already past.
By James W. Moore
If You Found This Useful
This newsletter exists because insurance executives need trusted, sourced analysis on AI — not vendor marketing dressed up as news. If a colleague would benefit from receiving it, forward this issue or share the link.
And if you haven’t downloaded The Independent Insurance Agent’s AI Playbook yet, it’s available free at insuranceindustry.ai/ai-playbook. More than 33,000 words of practical guidance, no cost, no catch.
Sources
- Mondaq / Taft Stettinius & Hollister: AI and Insurance Claims — Beware Fully Automated Decision Making
- Risk & Insurance: AI Risk Is Here — and Insurers Are Learning to Write the Rules
- Insurance Business America: How AllDigital Specialty Built an AI-First Insurer From the Ground Up
- Pulse 2.0: Pace — $46 Million Series B Raised for AI Insurance Operations Platform
- AllianceBernstein: How AI Is Changing the Landscape for Insurers
- Fortune: Microsoft AI Chief Gives It 18 Months — for All White-Collar Work to Be Automated by AI
AI Disclaimer: This content was created with assistance from artificial intelligence technology. While content is based on factual information from the source material, readers should verify all details directly with the respective sources before making business decisions.
