Mark Sendak is the CEO of a healthcare AI company serving community hospitals. He is a founder of Health AI Partnership and formerly worked at the Duke Institute for Health Innovation for over a decade.

The leader of a community hospital system in Kentucky recently told me, “I’ve exhausted our resources trying to keep up with AI”. Another leader in South Dakota wasn’t sold by the marketing of a well-known AI vendor. “They’re not building for us,” he explained while pointing to press releases that touted multi-million dollar contracts with prestigious academic medical centers. “They don’t understand we serve Native American populations in four of the five poorest counties in the United States.”

6000 hospitals in America need AI and other technology solutions. AI vendors are building for the 400 hospitals affiliated with academic medical centers, while community health systems’ needs are going ignored by startups. I spent the last decade of my career helping build an innovation group at one of these large academic medical centers – and I can tell you that this is the most significant distribution failure in healthcare.

Academic medical centers and community hospital systems both deliver care in hospital and clinic settings, but have a few major differences. Academic medical centers typically house a medical school and serve as the primary training site for medical students, residents, and fellows. They also have large research programs with significant external grant funding. Meanwhile, community hospitals do not typically house medical schools and have limited externally funded research.

When I started Vega Health last year, I knew from the start that we would not be focusing on the big-name logos.  We knew that community health systems still had needs that other startups were treating as an afterthought. Meanwhile, they represent a whopping 93% of American hospitals.

The vendor ecosystem continues to act as if the most advanced health systems and academic medical centers should be their core buying segment (or only buying segment). The reasons for it do make sense. These are big, prestigious systems with hospitals that top national rankings, and technology companies can ink contracts for many millions of dollars in a single year. These contracts also tend to be sticky, meaning they don’t churn, and it’s perceived as easier to raise capital off of them.

But there are also massive challenges. As a group of health system executives and Cadence chief medical officer Dr. Eve Cunningham recently discussed on a Second Opinion webinar alongside executives from the University of Chicago and Corewell Health, academic medical centers tend to be risk-averse, and getting a contract requires months (if not years) of work with legal, compliance, IT - not to mention aligning internal stakeholders. 

Other startups may think that locking in the big logos of prestigious academic medical centers translates to rapid sales versus the long tail of community health systems. My experience suggests otherwise. For more than a decade at the Duke Institute for Health Innovation, I watched leading researchers and innovators at places like Duke, Michigan Medicine, and Kaiser Permanente develop best-in-class AI solutions that failed to scale broadly to community hospital systems.

Adoption of AI in US hospitals remains sharply divided: 86% among system-affiliated hospitals versus just 37% among independent facilities. Rural hospitals similarly lag: only 56% report any AI integration compared to 81% of urban hospitals.

So why aren’t the best AI solutions penetrating the market of community hospital systems? Well, one reason is that no one is paying attention to this cohort. 

Investors and operators hear "community hospitals" and worry about small deal sizes and limited TAM. The age-old wisdom is that giant businesses cannot be built based on an accumulation of small contracts with tiny customers. But the numbers tell a different story.

Non-AMC hospitals are actually much larger than they might seem. This segment represents a $45+ billion IT market. With average IT budgets of $8-10 million per hospital across 5,600 facilities, and AI solutions capturing a growing share of IT budgets, there’s room for $10+ billion in annual revenue across AI solution vendors. These budgets may not be as large as the academic medical centers, but the gap is getting smaller. 

Our early experience selling to this market validates the potential. Initial deal sizes in the six and seven-figure range. Expansion conversations to rapidly increase the number of purchased AI solutions happen quickly – even pre-sale. Our first customer initially focused on three use cases for a 6-month pilot. By the time the deal closed and the contract was signed five months later, the scope had expanded to three years and six use cases. 

Another common misconception? Investors and operators fear that every community hospital will need something different, bloating service costs and transforming scalable software businesses into consultancies.

Again, our experience reveals a different story. Decision-making authority is far more diffuse across academic medical centers (and the departments within academic medical centers) than community hospital systems. It’s not uncommon to see every hospital leader and clinical service line leader with a different set of needs, and each will have their own P&L. Customization is actually far more of a problem at academic medical centers than anywhere else. 

At Duke, even when we built AI solutions that were scaled within our system, different hospitals within our umbrella system created different workflows. Our innovation team went to great lengths to align governance structures across departments and stakeholders.

Our experience working with community hospital systems is a stark contrast. There is strong centralized decision-making and standard processes across 6 hospitals at one site and 11 hospitals in another site. While there is configuration and adaptation required for each customer, the variability across hospitals is much less than academic medical centers.

Lastly, let’s talk about sales cycles. Notoriously long at academic medical centers, what are we now seeing on the ground at community hospital systems?

When I started Vega Health, I braced myself for 18 to 24-month sales cycles. We raised a seed round with the goal of closing our first customer within 12 months.

Six months later, we have two customers, and we can attest to the speed at which community hospitals can move. C-suite executives have a mandate to rapidly build capabilities in AI and demonstrate value in the form of improved outcomes with equal or less cost and staff burden. Contracting goes faster out of necessity, because community hospital systems know they can’t grow a portfolio of AI solutions on their own. 

There is also real financial pressure that is moving community health systems to act fast. Rural hospitals face 18% claim denial rates compared to 10% for urban facilities. More than 700 rural hospitals risk closure due to financial hardship. They can't afford to pilot and evaluate for two years. They need solutions that generate real efficiencies and cost savings that work this quarter. 

This purchasing behavior is fundamentally different from AMCs with internal research and innovation functions. Our sales conversations with AMCs must account for internally built tools and engage a diverse set of stakeholders: What can be done internally? How will an investment in an external solution compete with internal research efforts? How will an externally procured platform integrate with bespoke internally built products? 

The retention story will ultimately be about partnership and trust. When a community hospital finds a partner willing to do what it takes to make AI work in their environment, they don’t expect to switch vendors. Annual contract values will grow as the number of solutions used through a platform increases. 

Policy is creating urgency. CMS's WISeR model mandates AI/ML-enabled prior authorization for select procedures in six states. The $50 billion Rural Health Transformation Program explicitly funds technology-enabled solutions, including AI tools. These aren't grants for a bunch of free pilots. They're dollars that accelerate real adoption. 

The technology is commoditizing, and buyers are overwhelmed with options. We're seeing the first wave of healthcare AI companies built from day one for community hospitals. We aren’t pitching to Mayo Clinic or Massachusetts General Brigham. We’re starting by asking: What does a 200-bed community hospital actually need, and how do we deliver it profitably at scale? 

Some incumbents will try to serve this market. The winners will be whoever stops trying to make community hospitals look like academic medical centers and starts building for the market that actually exists: community hospitals with real problems, no patience for complexity, and a desperate need for solutions that work. 

For operators and investors who understand the difference, there’s a huge market opportunity hiding in plain sight.

Mark Sendak, MD, MPP, is the Co-Founder and CEO of Vega Health, a health AI
platform company that facilitates the safe, effective, and responsible use of AI at scale. He also co-leads Health AI Partnership, a learning collaborative to develop and
disseminate best practices that advance the safe, effective, and responsible use of AI
software within healthcare delivery organizations. Prior to founding Vega Health, Mark
spent over 10 years at the Duke Institute for Health Innovation (DIHI), where he led
interdisciplinary teams of data scientists, clinicians, and machine learning experts to
build technologies that solve real clinical problems. Together with the DIHI team, he
built tools to transform chronic disease management within an Accountable Care
Organization and detection and management of inpatient deterioration within hospitals. While at Duke, he created and led the DIHI Clinical Research & Innovation scholarship for 10 years, which equips medical students with the business and data science skills required to lead health care innovations.