Your weekly analysis of AI developments in insurance.
Iowa Just Made AI-Only Claim Denials Illegal. Other States Are Watching.
Iowa Governor Kim Reynolds signed legislation this week that eliminates prior authorization requirements for cancer screenings and, critically, restricts how health insurers can use AI in claims decisions. The bill, which takes effect July 1, prohibits insurance companies from denying a claim based solely on an AI recommendation. Human review is now a legal requirement when AI is involved in a denial.
The Iowa Hospital Association’s board chair, Shelly Russell, put the concern plainly: “One of the things that we have seen over time is that there has been some effort to only use AI in denials.” The new law closes that door. It also establishes firm deadlines for insurers to communicate with hospitals about claim status and requires clear clinical explanations when care is denied.
This is not an isolated state action. Legislative attention to AI in claims adjudication has been building across multiple states, and the Iowa bill follows a pattern: prior authorization abuse becomes visible, legislators respond, insurers adapt. The AI component adds a new dimension to that cycle because automated denial systems can operate at a scale that makes the pattern harder to detect before it becomes a legislative issue.
Why This Matters for Insurance:
The practical question for health carriers is not whether to use AI in claims workflows — that ship has sailed. The question is how to document AI’s role in a way that demonstrates human judgment is present and accountable. Carriers that have deployed AI broadly in denial workflows without clear human-in-the-loop controls now face a design problem: retrofitting governance is significantly more expensive than building it in from the start. Iowa’s law is narrow in scope but broad in signal. Expect similar language in other state legislative sessions before year-end.
The Society of Actuaries Survey Just Named AI the Top Emerging Risk for Insurance. The Reasoning Deserves Attention.
A new Society of Actuaries Emerging Risk Survey — the 19th edition of the study — found that insurance and financial services leaders across all sectors now rank AI as the top emerging risk for 2026 and the years ahead. The framing of the result is worth reading carefully, because it cuts against the usual pattern.
Near-term risk priorities remained diverse: life carriers focused on financial volatility, P&C carriers on extreme weather, consulting firms on AI. But when researchers asked all groups about what sits on the three-to-seven year horizon, they converged on the same answer: technology risk, specifically AI. The SOA characterized AI as “the current unknown on the horizon” — an acknowledgment that the risk profile of the technology is not yet well-mapped, even among the professionals whose job is to map it.
That convergence across carrier types and time horizons is the number that should concentrate minds. Individual executives can rationalize short-term priorities. But when actuaries, risk managers, and senior executives across life, P&C, and consulting all land on the same long-horizon concern, the signal is hard to dismiss.
Why This Matters for Insurance:
Carriers that have framed AI governance primarily as a compliance exercise may need to reframe it as a risk management exercise. The SOA language — “unknown on the horizon” — reflects the genuine challenge: the risk profile of AI in insurance decisions is still being discovered, often through failures rather than forecasts. Boards that are receiving AI update reports focused on deployment velocity and cost savings may not be getting the other half of the picture. The SOA survey suggests risk leadership is increasingly aware of the gap.
Outmarket AI Raised $17 Million to Automate Insurance Agency Workflows. The E&O Number Is the One to Watch.
Outmarket AI, a San Francisco-based platform built specifically for insurance brokerages, announced a $17 million Series A this week led by Permanent Capital Ventures, with participation from SignalFire, Fika Ventures, TTV Capital, and Dash Fund. The round brings total funding to $21.7 million. Notably, the investor group includes strategic capital from independent agency networks and senior industry executives — not just venture capital.
The platform is integrated natively with the major agency management systems: Applied Epic, AMS360, HawkSoft, and Nexsure. Its flagship product, Proposal Builder, converts a multi-hour manual process into minutes. The company reports over 250 brokerages are in production, with customers reporting up to a 65% reduction in errors and omissions through AI-assisted policy comparison and gap detection.
That E&O figure deserves scrutiny. A 65% reduction in E&O exposure is a significant claim, and it connects AI adoption directly to a liability metric that agency principals and their professional liability carriers track closely. If that number holds across a broader population of deployments, it provides an ROI argument for agency AI that goes beyond time savings and into risk transfer economics.
Why This Matters for Insurance:
The independent agency AI market is attracting serious capital, and Outmarket’s investor mix — venture firms plus agency networks plus industry executives — reflects a bet that the distribution layer is where AI adoption creates durable value. Carriers that rely on independent agents should be tracking what their distribution force is deploying. An agency that uses AI for policy comparison, gap detection, and proposal generation is operating at a different information level than one that doesn’t. That gap eventually surfaces in submission quality, account retention, and the quality of the risk conversations carriers have at renewal.
Insurance AI Needs Its Own Guardrails. A Practitioner from Arthur J. Gallagher Makes the Case.
Insurance Thought Leadership published a detailed framework this week from Rakesh More, program lead for AI at Arthur J. Gallagher, on why generic AI safety controls are insufficient for insurance deployments. The argument is both practical and regulatory, and it is more specific than most governance commentary the trade press produces.
The core problem More identifies is that standard AI safety tools address content toxicity, PII detection, and basic prompt injection defense. Those controls are necessary but do not address insurance-specific accuracy requirements. A response can clear a generic safety filter while still mischaracterizing a coverage provision, inventing an exclusion that does not exist, or providing incorrect claims guidance. The consequences are not just bad customer experiences — they are potential compliance events.
According to ACORD research cited in the piece, 77% of insurers now use AI somewhere in their operations, and early implementations have demonstrated claims processing time reductions of as much as 75%. The global AI in insurance market, valued at $4.6 billion in 2022, is projected to reach $79.9 billion by 2032. The scale makes the governance gap consequential: hallucination rates of 15-30% documented in general-domain LLMs do not become acceptable just because a system is widely deployed.
More’s implementation framework is organized around five guardrail types: input validation, dialog management, retrieval validation, execution controls, and output verification. The NAIC AI Model Bulletin’s five expectation areas — governance, transparency, risk management, auditability, and vendor oversight — map directly onto this architecture. The piece also notes that insurers directing 72% of AI spending to technology and only 28% to change management creates what he calls a “critical architecture mismatch.”
Why This Matters for Insurance:
The practitioner source matters here. This is not a vendor making the case for its own governance product, and it is not a regulator describing what it wants to see. It is the AI program lead at one of the world’s largest insurance brokerages describing what responsible deployment actually requires. The specific call-out on change management spend — 72% technology, 28% people and process — is a number worth testing against your own organization’s AI budget allocation. The frameworks that survive regulatory examination will be the ones that invested in both sides of that equation.
LexisNexis Launched an AI Property Risk Model That Scores Six Perils at the Individual Property Level.
LexisNexis Risk Solutions announced this week the launch of Location Intelligence for Home, a neural network-based predictive model for U.S. home insurers that scores individual properties across six perils: hail, wind, weather-related water, non-weather-related water, freeze, and collapse. The model is built on industry-wide claims data and integrates directly into LexisNexis’s existing Smart Selection platform, which means it fits into carrier underwriting workflows without requiring a separate integration project.
The non-weather water angle is the most operationally interesting detail. In 2025, non-weather water claims represented 24% of all claims — the single largest category. Weather-related water claims represented only 4%. Yet traditional property underwriting approaches have historically weighted weather exposure heavily and non-weather water exposure less systematically. A model that scores freeze risk, pipe failure risk, and non-weather water risk at the property level is addressing the peril that actually drives the most claim frequency, not the one that generates the most news coverage.
Why This Matters for Insurance:
Home insurers have spent the last several years repricing and retreating from coastal and wildfire-exposed markets. The next underwriting challenge is the interior risk — the water claim that doesn’t make the news but does make the combined ratio. Scoring that risk at the individual property level, rather than the ZIP code or county level, creates pricing precision that affects both adverse selection and renewal retention. Carriers that can see non-weather water risk at address-level granularity before binding have a materially different underwriting conversation than those that cannot. Expect this category of property intelligence to attract more investment as the weather-exposed repricing cycle matures.
The MIT Study on Automation and Wages Deserves More Attention Than It Got.
A paper published this week in the Quarterly Journal of Economics, co-authored by MIT Institute Professor Daron Acemoglu (who shared the 2024 Nobel Prize in economics), provides the most rigorous quantitative assessment yet of how U.S. companies have actually used automation since 1980 — and the findings complicate the standard narrative significantly.
The study found that rather than pursue maximum efficiency, U.S. firms have frequently deployed automation to target employees receiving a “wage premium” — workers earning more than comparable peers. This pattern is estimated to account for 52% of the growth in income inequality from 1980 to 2016, with about 10 percentage points derived specifically from replacing wage-premium workers. The inefficient targeting has offset 60-90% of the potential productivity gains from automation during the period.
Acemoglu’s framing is precise: “You can reduce costs while reducing productivity.” The historical record suggests that many firms have chosen exactly that tradeoff — cutting labor costs through automation in ways that improved short-term profitability but dampened long-term productivity growth.
Why This Matters for Insurance:
Insurance is a labor-intensive industry with significant concentrations of exactly the knowledge workers the study describes — experienced claims handlers, senior underwriters, and account managers earning wage premiums relative to their peer groups. The MIT findings suggest that how AI is deployed in insurance organizations will matter as much as whether it is deployed. Automation that targets experienced professionals to reduce payroll costs may deliver short-term expense ratio improvement while eroding the institutional knowledge that underlies underwriting quality and client retention. The frame most likely to survive long-term scrutiny is the one that uses AI to extend the capacity of experienced people, not replace them.
From the AI World: What Happened When Dario Amodei and Jamie Dimon Shared a Stage
On May 5, Anthropic hosted The Briefing: Financial Services, a livestreamed event from New York City that drew unusually substantive post-event coverage — including a detailed analysis from Info-Tech Research Group’s SoftwareReviews team that is worth reading in full.
Anthropic CEO Dario Amodei and JPMorgan CEO Jamie Dimon shared a stage to discuss AI in financial services. The event included a CIO panel featuring AIG CEO Peter Zaffino, Goldman Sachs CIO Marco Argenti, and JPMorgan CIO Lori Beer — a roster that signals how seriously the largest financial institutions are engaging with Anthropic’s enterprise positioning.
On the product side, Anthropic announced ten prebuilt agents for high-volume finance workflows, including an insurance claims workflow. The Microsoft 365 integration for Excel, PowerPoint, Word, and Outlook became generally available, with context carrying across all four applications. New data connectors include Verisk — a direct insurance relevance — alongside Moody’s, Dun & Bradstreet, Experian, and others.
The Info-Tech analysis identified three signals from the event worth tracking for insurance executives. First, Anthropic’s revenue has grown at 80x against an internal projection of 10x, meaning the vendor’s trajectory is substantially faster than even its own leadership anticipated. Second, Anthropic’s chief economist presented data showing AI is now involved in at least a quarter of tasks across roughly half of all U.S. jobs — up from one-third of that penetration a year ago. Third, the regulatory picture has shifted: supervisors are now watching AI model vendors directly, not just individual bank deployments. That pattern will extend to insurance as state and federal AI oversight matures.
The Verisk connector is the detail most directly relevant to insurance readers. A native integration between Claude’s agentic workflows and Verisk’s data infrastructure means carriers already running on Verisk data can potentially extend AI workflows without a separate integration project. That changes the timeline and cost structure for agentic AI deployment at carriers with existing Verisk relationships.
Why This Matters for Insurance:
The financial services briefing matters to insurance for the same reason the AIG earnings call mattered last week: it shows how the largest regulated-industry organizations are moving from pilot to infrastructure. Insurance is a more fragmented and more heavily regulated market than investment banking, but the deployment pattern — prebuilt vertical agents, native data connectors, governed agentic workflows — will follow. The question is timing, and events like this one compress it.
By James W. Moore
Sources
- KCRG: Governor Signs Bill Eliminating Need for Health Insurance Companies to Authorize Cancer Screenings
- InsuranceNewsNet: AI Emerges as the Biggest Risk for Financial Leaders in 2026
- PR Newswire: Outmarket AI Raises $17M Series A to Power the Intelligence Era of Insurance
- The Insurer: Insurance AI Platform Outmarket Raises $17 Million in Series A
- Insurance Thought Leadership: Insurance AI Requires Specialized Guardrails
- PR Newswire: LexisNexis Risk Solutions Launches AI-Driven Location Intelligence for U.S. Home Insurance Carriers
- MIT News: Study: Firms Often Use Automation to Control Certain Workers’ Wages
- Quarterly Journal of Economics: “Automation and Rent Dissipation: Implications for Wages, Inequality, and Productivity”
- Anthropic: The Briefing: Financial Services Virtual Event
- SoftwareReviews / Info-Tech Research Group: Anthropic’s “The Briefing: Financial Services” Event Was Different in the Best Way
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.
