Starr’s AI Strategy Runs Against the Insurance Hype Cycle
Fewer bots, fewer patents, and a sharper focus on claims, control, and regulation
Insurance’s AI divide is widening.
Most carriers said artificial intelligence was a strategic priority for 2025. Far fewer had moved it into production. One industry survey found that about 73% of commercial P&C CEOs list AI as a priority, while only ~7% report enterprise-wide deployment beyond pilots. The gap has become the defining feature of insurance AI adoption.
The divergence is visible in how firms sequence decisions. Some carriers moved quickly into customer-facing automation and patent filings. Others moved more slowly, focusing on governance and core operations first. Starr Companies, a global insurer and investment organization, has taken the latter approach, with a lower public AI profile but recent moves that suggest a deliberate operating-model shift rather than hesitation.
Starr Skipped the AI Showmanship
Starr’s first explicit signal came with the appointment of Yvonne Li as Chief Data and AI Officer in late 2024, consolidating data, analytics, and AI oversight under a single executive. The role matters in the current regulatory environment. U.S. insurance regulators, through the NAIC Model Bulletin on the Use of AI, now expect insurers to document, govern, and audit AI systems that affect consumers, pushing AI accountability out of innovation teams and into core management.
Starr’s other concrete move has been operational. In January 2026, the company said it was modernizing its property and specialty claims operations using Five Sigma’s AI-native claims platform, replacing legacy workflows rather than layering tools on top. Claims are the most economically consequential place to deploy AI. Consulting benchmarks estimate that claims handling represents 60–70% of an insurer’s operating costs, making it the largest controllable expense line.
The economics explain the focus. McKinsey estimates that up to 60% of claims activities could be automated by 2030, with potential cost reductions of roughly 30% when AI is embedded across the workflow rather than used as a point solution. For specialty insurers, where claims are lower frequency but higher severity, AI functions more as a triage and orchestration layer.
What Starr has not done is equally visible. The company does not appear among the largest filers of insurance AI patents. Data compiled by Insurance Journal shows that State Farm, USAA, and Allstate together account for roughly 77% of AI-related patents filed by insurers, concentrated in personal lines, telematics, and customer interaction technologies. Starr has also not publicly disclosed large-scale deployments of generative AI in customer communications or underwriting decisions.
The Risks Lurking Behind Insurance’s AI Boom
By contrast, some competitors moved quickly into high-visibility AI use. Allstate has publicly discussed using generative AI models to draft large volumes of claims communications, shifting adjusters toward review rather than authorship. Insurtechs like Lemonade automated first notice of loss and routine claims end-to-end, reporting faster settlement times for simple cases (en.wikipedia.org).
Underwriting has seen similar acceleration. Market analyses attribute ~25% improvements in underwriting efficiency and ~15% faster claims processing to AI adoption, driven by data ingestion, segmentation, and automated decision support. Specialist vendors have extended AI use into core risk decisions. ZestyAI’s property-risk models, for example, have received regulatory approval in more than 35 U.S. states, allowing insurers to use AI outputs directly in pricing and underwriting.
That progress has raised governance pressure. Multiple states have adopted versions of the NAIC’s AI bulletin, which requires insurers to maintain documentation, monitor outcomes, and address bias risks in AI systems that affect consumers. Regulators are also developing evaluation tools to inventory insurer AI use across underwriting, claims, and marketing.
The regulatory context matters for firms with investment exposure. Insurance groups have begun narrowing or excluding broad AI liability coverage, citing uncertainty around model behavior and downstream risk, a signal that AI’s risk profile is being reassessed across financial services. That environment favors approaches that prioritize auditability and control over speed.
Consulting research suggests that the next phase of insurance AI adoption will be decided less by experimentation and more by integration. McKinsey has argued that value accrues when AI is embedded into operating models rather than bolted onto legacy processes. The data supports the claim that many insurers are still early in that transition.
Starr’s approach fits that pattern. It does not demonstrate leadership by volume of AI announcements or patents. It demonstrates sequencing: executive ownership first, core cost center second, and public use cases later. Whether that restraint proves advantageous will depend on execution. What is already clear is that the insurance AI race is no longer about who deploys first. It is about who can scale systems that regulators, auditors, and balance sheets can tolerate.