By: Dan Goldsmith, Co-Founder & Partner, Proofpoint Capital
Everyone wants to know the same thing right now: as AI becomes pervasive, who is actually going to win in healthcare?
It’s the right question, but most people are asking it the wrong way.
Healthcare AI is no longer a science project. More than 70% of healthcare organizations say they are already pursuing or implementing generative AI. At the same time, capital is flooding the market: in the first half of 2025 alone, AI-enabled startups captured 62% of all U.S. digital health venture funding, and 9 of the 11 $100 million-plus digital health financings went to AI companies. The money is here. The urgency is real. But capital deployment is not the same thing as durable market ownership.
In every major technology shift – Internet, mobile, cloud – the biggest winners were not simply the first movers or the best-funded companies. They were the companies that became indispensable as the market reorganized around a new operating model.
Healthcare will be no different. But healthcare has its own rules. This is a market where trust matters more than demos, workflow matters more than model novelty, and measurable outcomes matter more than speed.
That is why I think the winners in healthcare’s AI era will not come from the extremes. Not the biggest incumbents by default. Not the flashiest AI-native startups by default. The best-positioned companies are often the ones in the middle: companies with real distribution, real customer trust, real domain context – and enough urgency to reinvent themselves before the market does it for them.
Three kinds of companies
Every market has three kinds of players.
The first is the large incumbent: Microsoft, Amazon, IBM, and the major established healthcare platforms. They start with obvious advantages – distribution, embedded workflows, trusted brands, installed bases, and the capital to keep adapting. In acute-care EHR, for example, Epic was the only vendor chosen by large health systems making go-forward enterprise decisions in 2024, and it posted its largest net hospital gain on record: 176 hospitals. That is what platform power looks like. And when incumbents decide a category matters, they can move quickly enough to flatten a lot of startups that thought they had more time.
But incumbents also face the innovator’s dilemma in its purest form. Their installed base is an asset until it becomes an anchor. AI lowers the cost of creating software, compresses the value of legacy technical moats, and exposes old architecture and accumulated tech debt faster than prior platform shifts did. Incumbents still have a head start—but AI is shortening the half-life of that advantage.
The second is the AI-native startup: fast, ambitious, technically sharp, and built for this moment. They have the benefit of starting with a clean sheet of paper. No legacy stack. No organizational antibodies. No fear of cannibalizing yesterday’s product.
But this is also where I see the most false positives.
Healthcare buyers are not handing out long-term contracts because a product feels magical in a demo. They are asking harder questions: Does it integrate cleanly into existing workflow? Does it reduce risk? Does it create measurable ROI? Does it earn trust fast enough to justify operational change? Those questions matter because healthcare is already deep in the experimentation phase. Among provider organizations, 30% report system-wide AI deployments, 22% are in implementation, and 40% are actively piloting solutions. Yet only 32% of surveyed buyers said startups have best-of-breed GenAI solutions superior to those from large tech incumbents. In other words, novelty alone does not carry the day.
That is the trap for the startup crowd: confusing speed of product development with speed of market trust.
The ambient-listening category is a good example. It has attracted enormous attention and serious capital. Suki announced a $70 million Series D in late 2024 and said it had raised $168 million to date by early 2025. Abridge says it now partners with more than 150 enterprise health systems and recently closed a new Series E round. That kind of traction is real. But the category is also showing how quickly features compress when platforms and distribution giants enter the field. Epic and Microsoft/Nuance have already expanded ambient documentation directly into core clinical workflow. In categories like this, capital can build speed—but not necessarily defensibility.
The third group – and the one I’m most interested in – is what I think of as the head-start companies.
These are companies that have already earned relevance in healthcare. They have customers, trust, real workflows, and referenceable outcomes. But they are not so large or so burdened by legacy systems that they cannot still move like challengers. They sit in the most interesting part of the market: credible enough to matter, but still hungry enough to change.
That position is more powerful than it looks.
McKinsey found that among healthcare organizations already implementing generative AI, 59% are partnering with third-party vendors to build customized solutions, while only 17% expect to buy off-the-shelf tools. That should tell us something important: healthcare buyers are not looking for generic AI. They are looking for partners who can help them operationalize AI safely inside the complexity of their own organizations.
That is why the middle matters.
But this group has its own trap: complacency. A company with growth, customers, and brand equity can easily convince itself that adding AI features is the same as becoming AI-native. It isn’t. Winning this era requires rethinking the business itself—product, implementation, customer success, operations, go-to-market, even org design. AI cannot just live in the roadmap. It has to live in the operating model.
The best companies in this middle lane will do something else that matters: they will pull their best customers into the innovation loop. In healthcare, co-development is not a nice-to-have. It is how new products become trustworthy enough to deploy. It is how you shorten the path from concept to referenceable value.
Innovation with empathy
There is another mistake I see all the time in healthcare AI: the assumption that “AI-native” is enough.
It isn’t.
Being AI-native is becoming table stakes. Being healthcare-native is still rare.
Healthcare is not a blank canvas waiting for Silicon Valley to redraw it. It is a live operating system—messy, regulated, understaffed, risk-sensitive, and mission-critical. If it were easy to fix, it would already be fixed. The companies that earn the right to lead here are the ones that respect what the market has already learned the hard way.
This is not just cultural. It is economic.
The CAQH Index found that the industry spends $89 billion annually on the administrative transaction categories it tracks, with $18.3 billion in savings opportunity still available through fuller automation. Providers account for the vast majority of that opportunity. That means the real prize in healthcare AI is not just impressive output—it is measurable operational relief. Time back. Labor saved. Denials reduced. Access improved. Staff burden lowered. In healthcare, the winners will be the companies that can prove they removed friction from the system, not just added intelligence to it.
That is also why security, privacy, and governance cannot be afterthoughts. McKinsey found that risk concerns are the top barrier to scaling generative AI in healthcare, followed by capability gaps, data and technology infrastructure, and proof of value. That is a very healthcare-specific hierarchy. The market is telling us plainly: if your product creates anxiety for compliance, IT, clinical leadership, or operations, you are not accelerating adoption—you are slowing it down.
What it actually takes to win
The companies that win markets like this are almost never the ones that raised the most money or wrote the most code. They are the ones that made themselves most critical to customer success.
I lived that lesson at Veeva. We did not build one of the most successful enterprise software businesses in history by trying to be everything to everyone. We built it by becoming indispensable to a specific set of customers in a complicated, regulated industry.
That is why I am bullish on companies that already have a real head start in healthcare, and why I’m especially interested in companies like Artera.
Artera was not imported into healthcare from somewhere else. It was built inside the category. It has more than a decade of experience, serves over 1,000 healthcare organizations, supports 2 billion patient communications annually, and recently surpassed $100 million in contracted annual recurring revenue. KLAS also ranked Artera #1 in Patient Communications for 2026, with an overall score of 89.8. Those are not vanity metrics. They are signals of relevance, trust, and installed workflow.
What makes that interesting to me is not simply that Artera has scale. It is that the company is trying to do the hard thing: combine healthcare-native trust with AI-native reinvention, while treating privacy and security as baseline requirements rather than future patches. That combination is rare. Startups often have agility without credibility. Incumbents often have credibility without agility. The companies with the best chance to define the next era are the ones that can earn both.
Healthcare’s AI era will not belong to whoever moves fastest. It will belong to whoever becomes hardest to replace.
That means owning workflow, not just features. Delivering outcomes, not just output. Earning trust, not just attention.
That is the bet I would make.
