Editor’s Note
Artificial intelligence is starting to look less like a technology project and more like institutional infrastructure.
The evidence this week comes from very different directions.
The United Arab Emirates is working on a framework for deciding which government tasks an AI agent may actually perform. Palantir hired former AIG chief executive Peter Zaffino to lead its global financial-services business. Researchers demonstrated that artificial intelligence can extract signs of serious heart disease from an ordinary electrocardiogram in less than two seconds. And Anthropic reportedly reached an annualized revenue run rate of $65 billion in July.
Those are not four versions of the same story.
But they point toward the same stage of development.
The difficult questions are increasingly becoming less about whether artificial intelligence works and more about what authority institutions give it, what information it can uncover, how deeply it becomes embedded in operations, and who builds the infrastructure around it.
For insurers, that is where AI starts getting interesting.
It is also where the consequences become harder to isolate as simply “technology.”
– James W. Moore, Editor-in-Chief
Palantir Just Hired an Insurance CEO
Peter Zaffino is moving from AIG to Palantir.
That sentence is probably more interesting than another announcement about an insurance company buying artificial-intelligence software.
Palantir announced September 2 that Zaffino, former chief executive and executive chairman of AIG, will become its Global Head of Financial Services beginning January 15, 2027. His responsibilities will include insurance companies, banks, asset managers, private-equity firms, and other financial institutions.
Zaffino is not simply an insurance executive Palantir recruited to give the company industry credibility.
The companies already know each other.
Palantir CEO Alex Karp said the companies had partnered to deploy Palantir products inside AIG. Zaffino, meanwhile, said financial institutions that lead in the future will be those building “durable AI infrastructure” today.
There is an important signal in the job itself.
Palantir is not hiring Zaffino to head insurance sales. It is putting a former chief executive of one of the world’s largest commercial insurers in charge of financial services globally.
That suggests Palantir sees financial services, including insurance, as an operating market substantial enough to justify industry leadership at the highest level.
It also reflects something happening more broadly across enterprise AI.
The competitive contest is moving beyond selling models or isolated AI applications. Increasingly, technology companies want to become part of the underlying decision and operating infrastructure of large institutions.
Insurance is an unusually demanding place to attempt that. The data is complicated. Authority is distributed. Decisions can have financial, contractual, and regulatory consequences. Existing systems are deeply embedded.
That may also be precisely why it is attractive.
Why it matters: Hiring a former AIG CEO to run financial services is a much stronger statement about Palantir’s ambitions than launching another insurance-specific AI feature. The companies positioning themselves to win enterprise AI may increasingly compete for institutional expertise as aggressively as they compete for technical talent.
The UAE Is Trying to Decide What an AI Agent May Decide
The United Arab Emirates has set an ambitious target: transition 50 percent of government sectors, services, and operations to agentic-AI models within two years.
The interesting part is no longer the target.
It is the classification problem underneath it.
The UAE Cabinet has established a governance framework assigning responsibilities across ministries and federal entities, while implementation work has begun mapping services, processes, and tasks to determine where agentic systems can be used. The government has identified autonomous execution and decision-making as explicit parts of the initiative.
More recent implementation work has moved directly into determining which tasks are appropriate for AI. The stated operating principle is straightforward: human leads, AI enables.
That sounds simple.
It is not.
Once an AI system can do more than retrieve information or recommend an action, somebody has to decide where its authority begins and ends.
Can it prepare a tax audit, or complete one?
Can it recommend procurement decisions, or approve them?
Can it assemble evidence for a government determination, or make the determination itself?
The same question is arriving in insurance.
An underwriting agent can collect information, evaluate guidelines, calculate indications, identify exceptions, recommend terms, and potentially bind coverage.
Those are not equivalent authorities simply because one technology can perform all of them.
The UAE project is particularly interesting because it is being forced to make those distinctions across an enormous portfolio of consequential government activity.
That may eventually make the resulting classification framework more important than the AI models themselves.
Why it matters: Agentic AI governance becomes much more concrete once organizations stop asking what AI can do and start specifying what it is allowed to do. Insurers facing the same transition may ultimately need classification systems built around authority, not merely around technology or use cases.
A Routine ECG May Contain More Risk Information Than We Thought
The electrocardiogram is more than a century old.
Artificial intelligence may have just changed how much information we can get from it.
Researchers led by Imperial College London have developed an AI system capable of identifying signs of heart failure and heart-valve disease from routine electrocardiogram data in less than two seconds.
The research, funded by the British Heart Foundation, trained the technology using 10.6 million ECGs and then tested it against tens of thousands of patient records. The system identified up to 81 percent of patients with reduced heart pumping function and up to 90 percent of patients with a common form of heart-valve disease.
The AI is not replacing a diagnostic echocardiogram.
It is finding signals in an existing test that humans generally cannot see, potentially allowing physicians to identify which patients should receive additional testing sooner.
That distinction is worth thinking about beyond medicine.
Insurance discussions about artificial intelligence often focus on obtaining more data.
AI may also increase the amount of usable information hidden inside data the industry already possesses.
A loss history, inspection image, recorded call, adjuster note, application, telematics stream, financial statement, or medical record may contain patterns that traditional analysis either could not identify or could not economically extract.
That potentially changes more than operational efficiency.
It changes the information available to a decision.
For life and health insurers, advances in medical detection could eventually affect underwriting information, mortality and morbidity assumptions, disease management, and the timing at which previously hidden conditions become observable.
More broadly, it demonstrates an important AI capability: improving risk visibility without necessarily creating a new source of data.
Why it matters: Some of AI’s greatest value to insurance may come not from collecting more information, but from extracting more meaning from information already being collected. Better visibility does not eliminate uncertainty, but it can change which uncertainty remains hidden.
Anthropic’s $65 Billion Number Is Getting Difficult to Call Experimental
Anthropic reportedly reached an annualized revenue run rate of approximately $65 billion at the end of July.
That number requires an important qualification.
It is a run rate, not $65 billion of annual revenue already earned. It extrapolates a recent revenue pace over a full year.
Even with that caveat, the trajectory is remarkable.
CNBC reported that Anthropic’s run rate had risen from roughly $9 billion at the end of 2025 to $47 billion in May and $65 billion by the end of July.
There is another caveat.
Recent reporting indicates that Anthropic and OpenAI may account for revenue differently, particularly revenue generated through cloud partners, making direct comparisons between their headline run-rate figures less straightforward than they appear.
But the precise comparison is not the most important point.
Scale is.
Only a few years ago, enterprise generative-AI spending largely consisted of pilots, experimentation, and relatively small software subscriptions.
A model company is now reporting an annualized revenue pace measured in tens of billions of dollars.
Whatever eventual public filings reveal about margins, accounting, infrastructure costs, or customer concentration, organizations are clearly spending real money to put these systems into production.
That has implications for insurance technology strategy.
AI infrastructure is moving from an emerging technology category toward a major enterprise technology market. Vendors will consolidate around it. Core platforms will integrate with it. Capital will continue flowing toward it. And insurers will increasingly have to make architectural decisions that may persist considerably longer than the current generation of models.
The models will change.
The infrastructure choices may be harder to unwind.
Why it matters: The significant number is not whether Anthropic’s run rate is exactly comparable with another AI company. It is that enterprise demand for AI has become large enough to create an infrastructure market of its own. Insurers should increasingly evaluate AI architecture with the same long-term discipline applied to other foundational technology decisions.
Sources
- Palantir Technologies, “Peter Zaffino to Join Palantir Technologies as Global Head of Financial Services”
- Business Insurance, “Zaffino to leave AIG, join Palantir”
- Dubai Media Office, “Mohammed bin Rashid reviews progress of project to deploy Agentic AI across 50% of government sectors within 2 years”
- UAE Cabinet / Dubai Media Office, federal framework for Agentic AI Project implementation
- AI News, “Agentic AI in government: the UAE decides what machines decide”
- British Heart Foundation, “AI finds hidden heart disease in under two seconds”
- The Guardian, “‘Superhuman’ AI tool spots heart disease in less than 2 seconds”
- CNBC, “Anthropic says annualized revenue climbed to $65 billion in July”
- Reuters, “Anthropic revenue run rate tops $65 billion, source says”
- Axios, “Anthropic and OpenAI’s revenue chasm, explained”
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.
