AI ate digital health (and what that means for fundraising)

24 healthcare founders reflect on their fundraising experience

This last year, I oversaw an independent study with two of my incredible MBA students, Kyra Gardner and Andy Wang (both CBS ‘26). They built out a database of healthcare AI startups and interviewed 24 founders, the findings from which became this piece.

---

AI is no longer a category. It is the category.

In Rock Health’s Q1 2026 funding report, AI disappeared as its own category. Not because the trend faded, but because it took over. Almost every digital health company now incorporates AI in some form. The category collapsed because the category won.

That collapse is the cleanest signal we have for what has happened to digital health fundraising. AI has flattened the market without clarifying it. Investors only want to invest in AI. But what they’re asking for seems to be a moving target.

For example, in just twelve months, the same investors told Eliana Berger two opposite things. At her seed: “AI agents are going to win the market. Why aren’t you building an agent?” At her Series A, less than a year later: “The data platform approach is the way to go.”

Investor consensus on what will win healthcare AI rewrote itself in the time it took her to ship a product. That whiplash exemplifies the state of digital health fundraising in 2026.

To understand what this looks like from inside, we interviewed 24 founders and CEOs of digital health companies who have been fundraising in the past twelve months. They told us what investors are asking, where the bar has moved, and what they wish they had known going in.

Digital health fundraising in 2026

To provide some context, two things are happening at once: funding for digital health is up 35% YoY after the post-pandemic slump. Yet the capital is concentrating in fewer companies.

Twelve mega deals accounted for 59% of all Q1 digital health funding, one of the highest concentrations Rock Health has ever recorded. According to HealthTech Alpha, the average deal size in digital health jumped to $60.5M in Q1 2026, up from $29.3M in all of 2025, while the median deal size more than doubled.

CB Insights reported the same pattern in Q1: average deal size climbed, total dollars climbed, but the number of companies sharing those dollars fell. And this isn’t unique to healthcare. Zoom out across all sectors, and the story is the same. In 2025, half of all US venture dollars flowed to just 0.05% of deals.

And then there’s the velocity story.

For the AI-native winners, the gap between rounds has compressed to a degree that would have been unthinkable just a few years ago. Abridge, OpenEvidence, Hippocratic AI, and Tennr all closed two to three funding rounds within a year or two.

For everyone else, the gap between rounds is stretching. According to Carta, the median time from seed to Series A now sits at 616 days, up from 420 days in 2021, a 44% increase. Timing between rounds for digital health startups overall is even longer. Many companies are stuck between a peak-era valuation they cannot grow into and a market that has moved on. The bifurcation is not just where the capital goes. It is how fast it gets there, and whether it gets there at all.

“Revenue does not equal revenue” and other lessons these founders learned

We asked founders what investors were asking in pitch meetings and during due diligence. We expected them to tell us about technical diligence. Model architecture. Hallucination risk. Data governance. That is not what we heard. Two questions dominated, and a third metric is reshaping how fast companies move between rounds.

Defensibility: will Epic eat your lunch?

The first concern is always defensibility. Investors keep asking variants of the same question: are you one model release away from being obsolete? They want to know if OpenAI or Epic will eat your lunch next time they announce a new feature.

“No one is getting rid of Epic. Epic will kill you.” — Anu Sharma, Millie

Punit Soni at Suki framed the same dynamic as the central selection criterion for who gets rewarded in this space. Companies either project a deep partnership with the system of record, figure out spaces where those systems do not play, or they become the system of record themselves.

Punit Soni, Founder & CEO of Suki, an AI-powered voice assistant designed to reduce administrative burden for clinicians.

Jesse Creighton at Heidi agrees with the Epic threat and believes the key is to build where Epic will not follow. His playbook has been to go towards areas that require regulation and compliance, like hardware, autonomous agents, Class II medical devices, FDA-approved prescribing pathways, or anything that takes years and a quality system rather than a model release.

Live ARR is the new bar

The rise of AI means it’s cheaper and faster than ever to start a company. Investors want to know what you’ve actually delivered, and how fast you’re growing. The adoption curve of the breakout AI-native companies has set a high bar. For instance, OpenEvidence, launched in 2023, is now used daily by 65% of US physicians. Google Ventures, one of their main investors, calls it the fastest-adopted clinical product since the iPhone. (You can hear from OpenEvidence co-founder and CTO Zack Ziegler on the Heart of Healthcare Podcast).

The broader market signal is just as clear: per Menlo Ventures’ State of AI in Healthcare, healthcare is adopting AI at 2.2x the pace of the broader economy, with 22% of healthcare organizations now deploying domain-specific AI tools, a 7x jump from 2024. Companies that cannot show a meaningful usage trajectory are being told to “come back later.”

“What are investors looking for? Adoption and workflow integration. ‘Does the AI work’ (for better and worse) - is proxied by, can you get anyone to adopt it?” — Narinder Singh, LookDeep

Jung Park at Parakeet Health described the shift as moving from theoretical ARR to live ARR. A year ago, founders could raise on what was possible. Now investors want receipts. Syam Palakurthy at SamaCare pushed the point further: not all revenue is equal. Enterprise willingness to spend on AI right now is higher than it has ever been, which means some dollars flowing in look like traction but are really a market paying for optionality. Investors are learning to tell the difference, and the founders who cannot explain which of their dollars are durable may get caught in the next repricing.

The metrics investors actually want to see are concrete:

• Daily active users as a share of licensed seats. Not how many you sold, but how many are actually using it regularly.

• Net Revenue Retention (NRR). Per Bessemer, a $30M ARR company raising at $1B+ should be showing 120%+ NRR; sub-100% gets flagged as fragile.

• Time-to-deploy. How fast can a health system go from signed contract to live product?

• Workflow capture rate. For AI scribes, the share of clinical notes captured. For prior auth tools, the share of submissions filed.

• ARR per FTE. Bessemer’s benchmark for truly AI-native companies is $500K–$1M+ per employee. $200–400K signals a SaaS business. Below that, you’re running a services company with an AI veneer.

So how much growth do investors actually expect? At Series B, founders are hearing 3–4x year-over-year growth. For companies commanding $1B+ valuations, Bessemer puts the bar at 6–10x.

“VCs are asking why it takes 3 years to reach $20 million in revenue when other companies are doing it in 6 months.” — Akash Magoon, Adonis

What founders wish investors knew

Eliana Berger, CEO at Joyful Health, is helping healthcare practices identify and reclaim lost revenue.

A couple of investor blind spots came up repeatedly across our 24 conversations.

The first is pattern matching from other industries. Investors keep importing benchmarks from outside healthcare and asking why the curves do not match. Punit Soni at Suki and Akash Magoon at Adonis both pointed to healthcare’s longer build arc: slower top-line growth but far stickier customers, with NRR routinely north of 135% in the best-performing companies. Even the healthcare AI “supernovas” growing 6–10x annually are doing so on top of years of compounding workflow integration.

“Healthcare investors are biased by the non-healthcare things happening around them in the industry. Your benchmark is not how other healthcare companies are performing—it’s non-healthcare companies like Cursor growing to $100 million ARR in a few months.” — Narinder Singh, LookDeep

The second blind spot is confusing the product wrapper for the product itself. Two companies can both be “AI” while doing entirely different things. Thomas Knox at VitVio described the misclassification problem directly: a competitor running a single 2D camera in a hospital room gets bucketed against a deeply integrated multimodal product because both qualify as “AI.” Unit economics tell the real story, and, per Bessemer, the clearest signal of whether a company is truly AI-native or a wrapper is ARR per employee.

The third blind spot is underweighting the infrastructure layer. What counts as a long-term moat shifted in real time over the past year. Syam Palakurthy at SamaCare gave a concrete example: prior authorization, often pitched as a clean technical AI win, isn’t a purely technical problem. Rather, it’s a coordination and incentive problem; the AMA’s 2026 survey work shows physicians still average 43 prior auth requests per week and 12 staff hours, and physicians increasingly worry that AI tools are accelerating denials, not approvals. You could solve prior authorization with 1990s technology if you could solve both the coordination and the incentives.

Allon Bloch, co-founder & CEO of K Health, has raised $420 million in funding. He recently visited us at Columbia Business School for a panel discussion on the "Business of AI & Healthcare."

If AI isn’t the moat, what is?

Let’s circle back to defensibility. Very few founders cited AI itself as their moat. Several explicitly say it cannot be one. Ariel Katz at H1 put it bluntly: anyone with $20 can buy OpenAI access. Narinder Singh at LookDeep was even more pointed when he told us, “if a new model release from OpenAI or Anthropic threatens your business, you are probably in the wrong business.”

So if AI itself is not a moat, what is?

Workflow embedding and switching costs

Workflow integration was the most cited (and most contested) moat across our 24 conversations. The logic is straightforward: healthcare runs on intricate, vertical-specific rules, and the companies that capture those rules first become very hard to displace. Alex Cohen capitalized on this at HelloPatient by building so many technical hooks into legacy systems of record that the integration depth itself is a technical moat.

Workflow embedding is only as durable as the infrastructure beneath it. AI scribes are the cautionary tale. They embedded deeply into clinical documentation, but then Epic launched its own scribe built on the EHR layer on which those startups depended, and offered this feature at a fraction of the price.

Jung Park (left) co-founded Parakeet Health, a healthcare-native AI platform that automates patient communications and call center workflows.

Brand and trust

Healthcare moves at the speed of trust, so it was not surprising to hear founders name brand and trust as their moat. Estelle Giraud at Trellis Health pointed out that “moat” was a word coined by Warren Buffett to describe brand power. Buffett has often emphasized that a strong brand builds a sustainable competitive advantage in consumers’ minds, driving long-term value and durable profits. If your NPS is high enough that customers would defend you, no one can dislodge you.

Data as a compounding asset

Prashant Samant at Akido Labs was the most specific founder we spoke to about what defensible data actually looks like inside an AI product. The training corpus must be real, and the reinforcement loop must be live. The AI-native winners are not winning on model architecture; they’re winning on proprietary, longitudinal, outcome-linked datasets that competitors cannot quickly reconstruct.

“Our product isn’t trained on synthetic or abstract data. It’s not Claude with a wrapper. It’s built on a RAG model that takes 10 million real patient case studies, and lives alongside our clinic. So we’re doing back testing with actual performance.” — Prashant Samant, Akido Labs

Distribution and domain depth

Jung Park at Parakeet Health argued that distribution is the hardest moat to copy, especially in winner-takes-all markets. Owning the channel to the customer is more durable than any technical edge. That maps to where Menlo Ventures’ State of AI in Healthcare sees the action: health systems lead all sub-segments at 27% AI adoption, with outpatient providers at 18% and payers at 14%. The companies that have already secured those buyer relationships are running with a structural advantage that pure technical ability can’t close on its own.

“Biggest moat for any company: go-to-market strategy. Owning a distribution channel equals a durable moat. Technology always converges. Everything else becomes commoditized.” — Jung Park, Parakeet Health

Regulation and hardware: the non-software moats

Every moat above can be built faster with enough capital and engineering talent. Regulatory clearance cannot. FDA review cycles, CMS rulemaking, and manufacturing certification run on external timelines that no model release or engineering hire can shortcut. This is most acute for hardware-bound products, where the device itself can be built faster with more capital, but its Class II or Class III pathway cannot. The pace at which regulators are formalizing AI-device pathways (CMS and the FDA unveiled the RAPID program for devices in April 2026) only widens the gap between companies that have already cleared those gates and those still negotiating with model providers. As Jesse Creighton of Heidi told us, “You can vibe code whatever you want, but you can’t become a Class II medical device in less than two minutes.”

Founder advice for raising in 2026

Brothers Aman (left) and Akash (right) Magoon started Adonis, a revenue intelligence and automation platform that has raised $95 million.

We asked founders for their advice to early-stage founders pitching healthcare AI today. Four themes ran through almost every answer.

1. Forget AI (really!)

Every founder we spoke to said some version of the same thing: forget the AI label. Investors are asking the fundamental questions: what is your business model, what problem are you solving, what is your traction, and why will you win against the 500 other pitches this year?

Almost every company is pitching AI, which means it's table stakes, not a differentiator. Rajit Kumar at SpinSci suggested taking AI out of your head entirely and asking whether you have honed in on a problem that is actually in the critical path. Anu Sharma at Millie made the same point from the operator side. Build something useful, and worry less about getting it funded than about getting it adopted.

Jesse Creighton at Heidi argued software has become the ephemeral on-ramp to something more durable: a data flywheel feeding a self-reinforcing model that competitors cannot replicate.

2. Pick revenue-driving over cost-saving

If you want to be positioned for adoption, follow the money. Health systems buy things that drive revenue or materially reduce costs. But founders report the biggest problem right now is not cost — it is provider shortages getting in the way of revenue capture. Eliana Berger at Joyful Health put the buyer logic plainly: revenue-driving workflows are where health systems look to adopt AI first because the ROI is unambiguous.

Picking the right workflow only gets you in the room; staying there is the harder part. Alex Cohen at HelloPatient added a hard line on enterprise readiness. In healthcare, the shortcut is the fastest way to lose a contract. You cannot slop code your way through patient-facing or front-door workflows.

3. Think beyond this round

Right now is the easiest time to raise capital that founders in healthcare AI may ever see. General sentiment around market uncertainty has founders prioritizing getting cash in hand, increasing execution pressure, and reducing the margin of error to grow into prospective valuations. That makes it the most dangerous time to overcommit. Anu Sharma at Millie warned that this round is not the round that matters — the next one is, and the one after that.

The intuitive answer in an uncertain market is to maximize runway. Some of the founders we spoke to are doing the opposite. Jung Park at Parakeet Health laid out the math: raise at $X today, and the next round needs to clear roughly 3x to avoid a down round. With investors now scrutinizing live ARR rather than narrative, that step up gets harder to clear. Akash Magoon at Adonis described actively choosing lower valuation offers for exactly this reason, even at Series C.

4. Do not fake it

Leah Sparks at Wildflower pointed to the obvious tell. Every company suddenly became AI-enabled, and investors can hear it. Prashant Samant at Akido Labs turned the same point into a stress test: if you cannot ship, you are not actually an AI founder. He has been vibe coding himself, which means anyone in the room can. The label without the output is not a moat.

Fierce Healthcare’s April 2026 coverage of the PHTI prior authorization study reinforces the point from the buyer side. Health systems are now actively reviewing whether their AI vendors are actually delivering the promised throughput or just adding a marketing layer on top of legacy software, and the early evidence on prior authorization specifically suggests several “AI” tools are driving costs up, not down.

Anu Sharma, founder and CEO of Millie, an AI-enabled maternity clinic, raised $12 million in series A funding last year.

The through line

If there is a through line across all 24 founder conversations, it is this: the founders who are navigating this market well are not the ones chasing investor consensus. They are the ones who have tuned it out. As Allon Bloch at K Health put it, “everybody’s selling their book.” The consensus has been wrong too many times, too fast, to be worth following.

The market will likely keep concentrating. The bar will keep moving. And the gap between the companies that get funded and the ones that do not will have less to do with the technology than with whether the product is being adopted, paid for, and if the founder can prove it.

One last thing worth saying: venture capital is a means to an end, not an end in itself. The measure of success should never be how much you raised, but whether any of it moved the needle on a system that desperately needs to change.

---

A thank you

This piece is based on founder interviews conducted in March and April 2026 by Kyra Gardner and Andy Wang as part of a Columbia Business School independent study supervised by Professor Halle Tecco.

Our deepest thanks to the founders who shared their time, candor, and hard-won perspective with us. This piece would not exist without them:

• Akash Magoon, Adonis

• Prashant Samant, Akido Labs

• Ganesh Padmanabhan, Autonomize

• Feng Niu, Evidently

• Kaitlin Christine, Gabbi

• Ariel Katz, H1

• David Korsunsky, HeadsUp

• Jesse Creighton, Heidi

• Alex Cohen, HelloPatient

• Jeremy Friese, Humata Health

• Eliana Berger, Joyful Health

• AJ Loiacono, Judi Health

• Allon Bloch, K Health

• Narinder Singh, LookDeep

• Anu Sharma, Millie

• Jung Park, Parakeet Health

• Syam Palakurthy, SamaCare

• Rajit Kumar, SpinSci

• Punit Soni, Suki

• Susan Sly, ThePause

• Estelle Giraud, Trellis Health

• Sunita Mohanty, Vibrant Practice (now Ultralight)

• Thomas Knox, VitVio

• Leah Sparks, Wildflower

Next up: the investor side

Founders had a lot to say about how investors are diligencing AI in healthcare right now. Investors have a lot to say about it, too. Part 2 of this series flips the lens and asks the people writing the checks what they are actually screening for, where they think the bar is heading, and which categories they will not touch. Subscribe to be the first to read!

Read more

• The 60+ Best Digital Health Investors

• The Tech-Enablement Playbook for Healthcare Services Startups

• Venture Math: How VCs Value Startups

• An Overview and Directory of Corporate Venture Capital (CVC) in Healthcare

• 35+ Celebrities Investing in Healthcare