Your weekly analysis of AI developments in insurance.
A Hospital CEO Said the Quiet Part Out Loud About Replacing Radiologists with AI. Insurance Should Be Listening.
The CEO of America’s largest public hospital system told a room full of peers that he is ready to replace radiologists with artificial intelligence. Mitchell H. Katz, who leads the 11-hospital NYC Health + Hospitals network, made the statement at a Crain’s New York Business panel on March 25. His proposal: let AI handle first reads on mammograms and other imaging, with radiologists stepping in only when the system flags something abnormal.
Katz is not a technologist making a sales pitch. He is an internal medicine physician running a system that serves over a million patients annually. His argument is economic. Radiologists have become significantly more expensive as imaging demand has risen, and AI systems are producing results that, for certain low-risk screening populations, are remarkably accurate. David Lubarsky, CEO of Westchester Medical Center Health Network, backed the point at the same panel, stating that the AI his system uses returns incorrect negative results on mammograms only about three times out of 10,000 for women not considered high risk.
The reaction from radiologists was swift and pointed. Mohammed Suhail, a San Diego-based radiologist, called the proposal dangerous, arguing that any attempt at AI-only reads would result in patient harm. The debate echoes earlier comments from Anthropic CEO Dario Amodei, who was widely criticized by radiologists for overstating AI’s current capabilities in the specialty.
Why This Matters for Insurance:
This is not a technology story. It is a liability story. If hospitals begin shifting to an AI-first imaging model, even with radiologist review of flagged cases, the professional liability landscape changes fundamentally. Who carries the malpractice exposure when a missed diagnosis originated from an AI first read that was never reviewed by a human? What happens to medical professional liability pricing when the standard of care itself is in flux?
For carriers writing med-mal and hospital professional liability, this debate is a preview of coverage questions that are about to become very real. For health insurers, AI-first screening could dramatically expand access and reduce costs, but it also creates new regulatory and litigation risk. And for every insurance executive watching the “will AI replace professionals” debate unfold in someone else’s industry, radiology is the most advanced case study of what that transition actually looks like in practice: messy, contested, and full of unresolved questions about accountability.
Mexico’s Largest Auto Insurer Scales Agentic AI for Claims. The Numbers Are Worth Paying Attention To.
Quálitas, one of Mexico’s leading auto insurance companies, announced that it has expanded its deployment of SoundHound AI’s agentic AI platform to handle end-to-end claims resolution. The partnership, which began in 2022 with basic conversational AI for high-volume inquiries, has now scaled to support approximately 100,000 monthly calls, a 150% increase from when the system first launched.
The operational numbers tell the story. SoundHound’s AI agents are now handling more than 74% of car assistance requests end-to-end, over two-thirds of partial theft claims, more than three-quarters of broken glass claims, and over 80% of interactions where a policy number needs to be captured. These are not simple FAQ deflections. They are substantive insurance transactions that previously required human agents.
Why This Matters:
Quálitas represents something the insurance industry has been waiting to see: a carrier that started with a modest AI pilot four years ago and has now reached a scale where the technology is handling the majority of certain claim types autonomously. The progression from conversational AI to agentic AI is significant. The earlier system could answer questions. The current system can navigate multi-step processes, make contextual decisions, and resolve claims without human intervention.
For carriers watching the agentic AI space, the Quálitas deployment offers a realistic timeline for what scaling actually looks like. It took four years to go from pilot to this level of autonomous handling, and the insurer is still keeping human agents available for complex or sensitive interactions. That is a more honest picture than the “deploy AI and transform overnight” narrative that dominates vendor marketing.
Coforge and Solstice Target the P&C Core System Problem with Agentic AI
Coforge, a global IT services firm with deep insurance vertical expertise, announced a strategic partnership with Solstice Innovations to accelerate AI-led modernization of core insurance technology for P&C carriers. The collaboration combines Coforge’s Forge-X agentic AI delivery platform with Solstice’s Equinox core insurance system, which unifies policy, billing, claims, and agency management in a single configurable environment.
Under the agreement, Coforge will establish a dedicated Center of Excellence to support client onboarding, system integration, migration, quality engineering, and business process transformation. The Equinox platform spans multiple lines including NFIP flood, private flood, homeowners, and broader P&C products.
Why This Matters:
Core system modernization has been the insurance industry’s longest-running headache. Legacy platforms are deeply embedded in carrier operations, and replacing them is expensive, slow, and risky. What makes the Coforge-Solstice announcement different from the typical core system press release is the explicit application of agentic AI to the migration process itself. Rather than just building a modern destination platform, the partnership is using AI agents to accelerate the transition: automating data migration, testing, and integration tasks that have historically consumed years of IT resources.
For mid-sized P&C carriers that have been stuck on aging platforms but lack the budget for a multi-year, nine-figure transformation, this is worth watching. The promise is faster, lower-risk modernization. The proof will be in the first deployments.
Rosella Raises $3.7 Million to Automate Commercial Insurance Brokerage
A new AI-native commercial insurance brokerage called Rosella raised $3.7 million in pre-seed funding led by Peak XV Partners and Intact Private Capital. The company is targeting the workflows that consume the most time in small and mid-market commercial insurance: submissions across multiple carrier portals, policy comparison, coverage gap analysis, and servicing.
Rosella’s platform includes AI document intelligence for comparing policies and flagging coverage differences, a multi-portal submission agent that automates the process of submitting risks across more than 100 carrier portals, and AI-assisted phone support for live customer calls. The company says certificates of insurance that previously took 30 minutes can now be generated in under two minutes.
Why This Matters:
The commercial insurance submission process is one of the most labor-intensive workflows in the industry, and Rosella is attacking it directly. Any producer or account manager who has manually entered the same risk information into multiple carrier portals understands the problem. The fact that Intact Private Capital, an arm of one of Canada’s largest P&C carriers, is among the investors suggests that established carriers see value in this approach, not just venture capitalists.
The phrase from Rosella’s co-founder that stands out: the real prize is not chatbots but browser agents that can navigate a hundred carrier portals, each one different, each one changing daily. That is a specific, practical description of what agentic AI actually means in an insurance context.
California’s AI Procurement Executive Order Has Insurance Implications That Go Beyond State Contracts
On March 30, California Governor Gavin Newsom signed Executive Order N-5-26, establishing governing principles for the procurement and deployment of generative AI across California state government. The order directs multiple agencies to develop new vendor certification requirements, contractor responsibility reforms, and guidance on responsible AI adoption, with most deliverables due within 120 days.
The vendor certification requirements are notable. Companies contracting with California state agencies may be required to attest to and explain their AI policies around governance measures to reduce harmful bias, protections against violations of civil rights and liberties, and safeguards against the generation of illegal content.
The order also includes a provision directing the state’s CISO to review federal supply chain risk designations of technology companies and, if those designations are found to be improper, to issue guidance allowing state agencies to continue procurement. This follows the Pentagon’s recent designation of Anthropic as a “supply chain risk,” a designation that was subsequently enjoined by a federal court.
Why This Matters:
California has a long history of setting standards that the rest of the country eventually follows, from emissions regulations to consumer privacy law. This executive order explicitly states that public procurement is a powerful tool for shaping market behavior. For insurers that contract with California agencies, the new certification requirements could impose compliance obligations around AI governance, bias mitigation, and civil liberties protections.
More broadly, the order signals what AI vendor due diligence is going to look like going forward. The certification framework California develops over the next 120 days will likely influence commercial contracting requirements as well. Carriers and agencies already building AI governance documentation will be ahead of the curve. Those that are not should start.
Marc Andreessen Says AI Layoffs Are a Smokescreen. Oracle’s Layoffs Suggest It Is More Complicated Than That.
Two stories from this past week sit in direct tension with each other, and the insurance industry should be paying attention to both.
First, venture capitalist Marc Andreessen told the 20VC podcast that AI-driven layoffs are largely fictional. His argument: most large companies are overstaffed by 25% to 75% after pandemic-era hiring sprees, and AI has become the convenient excuse to clean house. He called it the “silver bullet excuse” and pushed back against the idea that the technology is actually sophisticated enough to replace human workers at scale.
Then, within days, Oracle began cutting thousands of jobs across its global workforce of 162,000. The reason: a cash crunch driven by the company’s massive investment in AI data center infrastructure. Oracle has committed to spending at least $50 billion this year on AI buildout, and the financial pressure of that commitment is forcing cost reductions elsewhere. The company’s stock has fallen roughly 25% this year. Reports suggest the cuts could reach 10,000 or more positions.
Why This Matters:
The insurance industry is both an observer and a participant in this dynamic. As underwriters of D&O, employment practices, and workforce-related risks, carriers need to understand whether AI layoffs represent genuine technological displacement, post-pandemic correction, or financial rebalancing to fund AI investments. The answer affects how you assess management liability, employment practices claims, and even workers’ compensation exposure as companies restructure.
The Oracle situation is particularly instructive because it represents a third category that neither the “AI is replacing everyone” nor the “AI is just an excuse” narratives capture well: companies cutting jobs not because AI replaced those workers, but because funding AI infrastructure consumed the capital that would have otherwise supported those positions. For insurance executives evaluating their own AI investment timelines, Oracle is a cautionary case study in the financial pressure that large-scale AI commitments can create.
A Federal AI Tool That Forecasts Drought 90 Days Out. Here Is Why Insurers Should Care.
The U.S. Geological Survey released River DroughtCast, a machine learning tool that forecasts streamflow drought conditions up to 90 days in advance at more than 3,000 locations nationwide. The system was trained on data from USGS streamgages, some with more than a century of continuous records, and it uses that historical data to predict when rivers and streams will drop to abnormally low levels.
The tool is most reliable in the first four to six weeks. For the first week, it correctly predicts the onset of severe or extreme drought conditions approximately 75% of the time. That accuracy drops to around 55% by week 13. All forecasts include confidence estimates so users can assess reliability.
Streamflow drought is distinct from the more familiar meteorological drought defined by lack of rainfall. Factors like soil moisture, snowpack, and groundwater all influence how dry conditions translate into reduced river flows, making streamflow drought particularly difficult to predict but critical for water resource planning.
Why This Matters:
For insurers writing agricultural coverage, crop insurance, or parametric products tied to water availability, a 90-day drought forecast changes the information landscape significantly. Farmers could adjust planting decisions. Municipal water managers could implement conservation measures earlier. And insurers could incorporate forward-looking drought probability into pricing and exposure management rather than relying solely on historical loss data.
The broader point for the industry is that government agencies are quietly building AI tools that could reshape how environmental risks are assessed and priced. River DroughtCast is built on publicly available data from thousands of gauging stations with decades of records. That is exactly the kind of high-quality, structured dataset that machine learning excels at interpreting. Expect more tools like this from USGS, NOAA, and other federal science agencies, and expect them to eventually feed into the models that insurers use to price climate-exposed risk.
Sources
- Radiology Business: CEO of America’s largest public hospital system says he’s ready to replace radiologists with AI
- GlobeNewsWire: Quálitas Scales Agentic AI for End-to-End Claims Resolution With SoundHound AI
- BusinessWire: Coforge and Solstice Innovations Partner to Redefine P&C Insurance Modernization Through Agentic AI
- FintechNews: Rosella Raises $3.7M to Automate US Commercial Insurance Brokerage
- Alston & Bird: California Jumps into AI Procurement with State Governing Principles in an Executive Order
- Fortune: Marc Andreessen says AI layoffs are a farce
- CNBC: Oracle cutting thousands in latest layoff round as company continues to ramp AI spending
- USGS: New AI Tool Forecasts Drought 90 Days Ahead Nationwide
AI Disclaimer: This blog post was created with assistance from artificial intelligence technology. While the content is based on factual information from the source material, readers should verify all details, pricing, and features directly with the respective AI tool providers before making business decisions. AI-generated content may not reflect the most current information, and individual results may vary. Always conduct your own research and due diligence before relying on information contained on this site.
