The End of the Healthcare Data Intermediary
Why AI is making analytics vendors obsolete.
For years, a specific breed of healthcare company has thrived in the fertile ground between self-insured employers and their own claims data. These analytics and data companies built entire business models on the premise of acquiring claims data from administrators, crunching the numbers, and returning insights and visualizations to the employer for a hefty fee.
It was a lucrative model when data analysis required specialized software and armies of data scientists. That era is ending, and the impending shift is going to create havoc in the venture capital world.
The premise driving this extinction event is straightforward. First, the employer owns their healthcare data. Second, claims administrators are obligated to deliver that data to them. Third, and most importantly, today’s unbounded artificial intelligence tools are every bit as good at providing insights and building revealing visuals on the fly as the specialized vendors charging a premium for the same service.
The AI equalizer
The democratization of data analysis is playing out in countless ways across corporate America. Today, there is virtually nothing a so-called “healthcare data company” can do with AI tools that a self-insured employer cannot easily do internally.
For companies that live in Microsoft Excel, Copilot is seamlessly bridging the gap between raw data and actionable insight. For those operating within the Google ecosystem, Gemini in Sheets is performing the exact same function. Anthropic’s tools, particularly the Claude models integrated across various platforms, are proving highly capable of parsing complex datasets and generating high-level strategic analyses in seconds.
If an HR executive or CFO wants to understand where their healthcare spend is leaking, identify utilization trends, or visualize pharmacy costs, they no longer need to wait on a third-party vendor’s proprietary dashboard. They can query their own data in plain English and get the answer instantly.
The value is in ownership, not analysis
The fundamental economics of healthcare data have flipped. The value is no longer in the analysis—it is purely in the ownership of the data itself.
We are seeing this reality borne out in recent technological testing. There is a growing consensus that unbounded AI models (general-purpose large language models) consistently outperform build-for-purpose AI in most cases where the underlying information is publicly available or structurally standard.
The specialized “medical AI” tools that flooded the market recently have proven largely unimpressive. In many instances, they perform less well than their unbounded counterparts. The only scenario where a fit-for-purpose AI tool maintains a competitive moat is when it sits on top of a highly proprietary, inaccessible dataset. But in the world of self-funded employer claims, the employer already owns the proprietary dataset. The moat is gone.
The coming reckoning
This reality has not fully set in for the broader market, but the clock is ticking. Over the next couple of years, we are going to witness a reckoning.
Investors have poured billions into startups attempting to build fit-for-purpose healthcare analytics tools. But those that rely on public data are already in the hands of customers at no charge. And those that rely on proprietary data are only of value to the holders of the data who can do the rest themselves.
The future of healthcare analytics doesn’t belong to the middlemen building thin wrappers around large language models. It belongs to the employers who control the raw data, and who now hold the tools to unlock its value themselves.