The leapfrog opportunity in healthcare AI
The healthcare industry has been a severe laggard in the adoption of tech (software in particular): it has consistently spent less than half on IT/software spend as a percentage of revenue versus peer industries, and workflows are primarily dominated by manual processes and legacy tools like faxes and phone calls. You don’t see anything close to the equivalent of the likes of Salesforce, Slack, JIRA, Notion dominating the app layer in the way workers do their jobs, as you do in other industries.
Whereas we historically have viewed the low adoption of software as a liability, we now view that as an asset—we don’t face the same sunk cost bias that might be slowing other industries down from taking full advantage of the latest AI innovations. Other industries have spent hundreds of billions on enterprise software apps and are now faced with the decision of whether to rip those last-gen tools out and effectively spend additional billions on the modern AI tools that are now emerging.
In healthcare, the question is simply—should I throw more bodies and fax machines at the problem, or should I jump to adopt AI without the baggage of last-generation workflow apps getting in my way?
Scaling supply to meet demand
Healthcare is dealing with the mother of all staffing crises—we’re short over 100,000 doctors and nurses relative to the level of (rapidly growing) demand for clinical services that is projected in the next 5 years. Medical care is only getting more and more complex—the steady beat of breakthroughs in diagnostics, continuous monitoring, and miracle drugs means clinicians are dealing with increasing information overload. They need new tools to be able to synthesize highly complex data sets, in real-time, at the point of care, to inform highly consequential decisions about how to manage populations of patients who are living longer with more complex diseases and behavioral issues.
The scarcest asset in healthcare is clinical judgment, which today only exists in the form of human doctors and nurses. So one of the most profound challenges we face in healthcare is how to scale clinical judgment beyond the clinicians we have, and beyond the walls of the hospitals and doctors’ offices we have, to make it accessible to all who need it, when they need it. Relatedly, for the doctors and nurses we DO have in the system, how do we ensure each one of them can perform at the same level of exceptional performance as the best of their peers in the world?
Administrative AI for automating the backend of healthcare is critical for reducing the overhead of care delivery against these goals. Clinical AI products are also uniquely positioned to address these opportunities—though they are amongst the hardest types of products to build, because of the high stakes nature of what they do, and the high stakes nature of the environments into which they must be embedded and adopted.
Regulatory rails for clinical AI
Along those lines, another reason healthcare is one of the industries best poised to take advantage of AI is that it already has well-established regulatory rails for approving AI products for use in the real-world clinical setting. The FDA has already approved hundreds of clinical AI products for use in the wild, and is developing frameworks for upgrading those processes to address the latest innovations in ML and generative AI.
The rite of passage that these regulatory processes represent mean that only the companies with the most clinically rigorous products and development methodologies will make it to market, which in turns means there’s a higher barrier to entry, but also a stronger moat for those that make it through.
The size of the prize
When describing the opportunity for tech in healthcare, everyone always talks about how it’s a massive $4 trillion+ industry. But the vast majority of that $4 trillion is services (human labor) spend, not tech spend; and per the earlier point, it’s been very hard for enterprise software companies to penetrate that morass and capture much value in that overall pie (IT budgets are only 3.5% of revenue in healthcare, which is less than half of what it is in the financial services industry).
But now, AI tools are getting good enough (and evolving rapidly enough) that organizations may view them more as “AI staff” than as software products. We’re not just talking about disrupting the market of enterprise software, which is on the order of tens of billions of dollars, but we’re talking about disrupting the market of services, which is on the order of trillions. So in that sense, the scale of opportunity is 1-2 orders of magnitude larger than historical software opportunities, which is reflected in the quantum of capital and valuations going into companies executing this strategy.
So whether you’re building a challenger clinic, an infrastructure company that is rebuilding the guts of how our healthcare system operates, or a new type of payment model or insurance—we believe there has been no better time to start a company in this space than now, and not the least because of the tectonic shift that AI represents in how companies can be built, scaled, and brought to market in healthcare.
