%20(1).png)
Beyond Subscription: The Business Model Case for AI-Native Vertical EHRs
In Part 1 of this series, we laid out why we believe AI-native, vertical EHRs could represent one of the more interesting bets in healthcare software right now. We described the structural dynamics that make certain specialties worth targeting: provider independence, workflow complexity, administrative burden, and competitive fragmentation.
We closed with a question we've been spending time on since: does the EHR market need business model innovation to be truly venture scale?
The “SaaSpocalypse” has made this question hard to ignore. If the terminal value of the world's largest software companies is being written down, verticalized ambulatory EHR software is not immune.
At its core, the "SaaSpocalypse" debate is a structural one: when AI can automate the workflows per-seat software was built to support, does the subscription model hold? For vertical EHRs, the question cuts even deeper. These platforms sit on top of enormous clinical context and currently do almost nothing with it. AI doesn't just put pressure on the subscription model. It opens the door to a better one, priced on consumption rather than seats.
That pressure has direct implications for the incumbents. The last wave of EHR PE (Thoma Bravo's acquisition of NextGen, Warburg's backing of ModMed, Francisco Partners' ownership of AdvancedMD) was underwritten on a familiar thesis: sticky subscriptions, low churn, customers too entrenched to migrate. If AI-native entrants show up with better software at a cheaper price and a business model that makes money elsewhere, how much of that stickiness really holds?
The EHR business model status quo
Most vertical EHRs today charge per provider, per seat, or as a flat monthly fee. It's a clean model: predictable, easy to price, easy to negotiate. But we believe it also dramatically underprices the value that an EHR, sitting at the center of a practice's workflow, is actually positioned to deliver.
EHRs capture an immense amount of data: every prescription decision, every diagnosis code, every prior auth submission, every referral, every lab order, every payer interaction, every billing event. And yet they do almost nothing with it, which is reflected in their pricing. Ambulatory EHRs capture 1-2% of the revenue a practice generates. The data live in the EHR, but the actions and labor happen elsewhere.
A pure subscription specialty EHR runs $150 to $1,500 per provider per month depending on specialty and practice size. A mid-sized gastroenterology practice with 10-15 providers might net a $60-100K ACV for a company like ModMed, before you layer in any RCM take rate. Not nothing, but modest given how central the software is to the practice. Even with RCM layered in, CAC is high and implementation timelines are long, which is a big part of why this category has been hard to build at venture pace. This pressure has pushed operators in this category to rethink how an EHR should actually make money.
Moving past SaaS in EHRs
Expanding beyond pure SaaS isn't a new idea. Flatiron Health, athenaHealth, and Practice Fusion have all tried it, each with a different monetization approach, and each with varying degrees of success.
Flatiron Health’s vision was that pairing deep specialty ontology with a data asset would deliver outsized value. Flatiron bought an oncology EHR, Altos Solutions, that enabled them to structure real-world clinical data that could be sold to pharma for clinical trial optimization, drug development R&D, and regulatory support. They translated messy, unstructured notes into curated, analyzable datasets. The EHR was the entry point; the curated oncology dataset was the business.
Roche acquired Flatiron in 2018 for $1.9B, yet the acquisition undermined the business model. Despite Flatiron being a standalone business, Roche’s competitors became wary of working with a data platform that was associated with a key competitor. Since then, Roche has sold different parts of the Flatiron business for a fraction of the sale price.
Although this data platform approach could theoretically work in other specialty markets, getting past this structural problem post-acquisition would be key to getting excited here.
Athena built its business around a different insight: if you process the revenue cycle for a practice, you should share in the outcomes, not just charge for the software. Although athena is now privately held by Hellman & Friedman, its public market history tells a clarifying story. The business launched in 2000 not as an EHR but as a cloud-based revenue cycle company. In fact, their EHR product didn't come until 2006. When you look at the public filings, the revenue composition reflects that origin: the overwhelming majority of revenue came from RCM and a percentage of collections, not software subscriptions. Athena is widely categorized as an EHR company, but the business model is fundamentally different from a pure play SaaS company. The EHR was the wedge, while RCM was the business. That's a large part of why $17B was a reasonable price for a company that was nominally "just an EHR."
It is key to note that while athena pioneered this model, most large specialty EHRs now bundle RCM services very tightly with their EHR product including ModMed, NexTech, eClinicalWorks and others. So even in a world of SaaS disruption, these ambulatory EHRs have more revenue durability than most people believe.
Practice Fusion took the most aggressive swing, offering their EHR for free. At its peak, it was one of the largest EHR networks in the country by provider count — valued at around $1.5B in 2016 — built entirely on a free model. Revenue came primarily from advertising, de-identified data sales, and clinical decision support. The latter is where the company got into trouble, after accepting payment from a pharmaceutical company in exchange for encouraging physicians to increase prescriptions for extended-release opioids. They ultimately settled with the DOJ for $145M and then Allscripts acquired the company in 2018 for $100M. Legal issues aside, the underlying hypothesis and vision was compelling: when you have access to a broad clinical network, the monetization surface is substantially larger than a seat license. We’ve seen a similar thesis play out more successfully with Doximity and now OpenEvidence.
So are there opportunities for innovation?
The through-line across these examples is that the business model follows the product. In Part 1, we laid out three archetypes for companies looking to unseat the incumbents.
- Action layers that sit on top of EHRs. This is where most venture dollars have flowed to date. Automate workflows, integrate with the existing EHR, prove ROI, expand. The catch is that you're building on a platform you don't own, and the incumbent can pull the rug whenever they decide to
- Modern systems of record. The clean-sheet rebuild: better data models, modern architecture, native automation from day one. In theory, owning the system of record unlocks all the same business model optionality as a full vertical OS, but the timeline to get there is brutal
- Full vertical operating systems. This is where the business model fully unlocks. These platforms combine the system of record with end-to-end workflow execution and layer in revenue streams across RCM, prior auth, specialty pharma routing, diagnostics, and data. Companies like Ease Health appear to be taking this approach. The EHR subscription isn’t the prize here, but rather a way to lock in the customer and capture all your value elsewhere
Of the three, the full vertical OS is the one we keep coming back to. It's the only archetype that holds up across the dimensions we care most about: AI leverage, long-term defensibility, and TAM expansion.
AI-driven consumption business models
The historical examples point to the same conclusion that stand-alone seat-based pricing is not the ideal business model for EHRs. Instead, Flatiron, athena, and Practice Fusion each found their way to a consumption-driven business model. Most notable is athena given they structurally shifted how the entire ambulatory EHR market operated and pushed companies deeper into the world of RCM.
So does AI create a new structural business model shift in this market?
We believe it does across two dimensions: 1. Improving the economics of already existing consumption models, 2. Enabling new consumption models to more easily emerge.
Improving existing consumption models:
RCM
RCM is where AI has the clearest path to value in this stack. The AI RCM market has already attracted roughly $1B in cumulative venture funding, with companies like Akasa, Cohere Health, and Adonis growing aggressively. The story is simple: AI reduces human labor across coding, billing, denial management, and appeals while improving collection rates. Today, athena takes 4-8% of billings for their services and with AI properly implemented, their profit margins should increase in a meaningful way.
In an ambulatory market where the dominant EHR doesn't provide RCM services, an AI-native competitor can offer a better EHR and undercut the standalone RCM vendor at the same time. The combined product is stronger than either piece sold on its own. But in markets where the incumbent EHR already bundles RCM, that wedge is closed, and an AI-native challenger has to find other consumption streams to compete on.
Enabling new consumption models:
Practice management
Many EHRs already layer in practice management capabilities, yet none have taken the step to truly “own” the administrative function and labor associated with it. We believe that’s where the next wave of value will be captured, and it has two dimensions: the labor that keeps the practice running and the analytics that grow it.
On the labor side, AI agents are already automating intake, scheduling, insurance eligibility, and other workflows that today are human-intensive and largely decoupled from care delivery. But the shift is not just in automation, it's in how these workflows are monetized. SaaS monetization has followed a clear arc: first, companies charged for software. Then the smartest companies realized they could charge for payments. Now, the next frontier is charging for labor. This is not feature expansion. It's a structural shift in how software companies will capture value. Rather than selling a seat license, companies are increasingly monetizing on work completed: per intake processed, per appointment scheduled, per eligibility check verified. In many cases, they are effectively becoming the staffing layer themselves, with a pricing model that directly maps to labor displacement and throughput.
On the intelligence side, the same system that captures every marketing touch, scheduling event, prescribing decision, and claim outcome is uniquely positioned to turn that exhaust into operational insight: CAC by referral source, LTV per patient, case acceptance rates by provider, the unit economics of an MA hire. Today, most practices either do this in spreadsheets or fly blind. Incumbent EHRs surface dashboards that technically exist but rarely answer the questions a practice owner actually needs to run their business. An AI-native system of record closes that loop natively, which is what it really means to run the practice and not just record it.
Folding staffing and analytics into the EHR is not just product expansion. It's the logical end state of this monetization shift. The EHR that moves first to own both layers will not just be a system of record. It will be the system that actually runs the practice end-to-end.
Clinical decisioning & workflows
On the clinical side, incumbents have conspicuously ignored assisting in clinical decisioning and the workflows that sit alongside these decisions. Instead, companies like Navina aggregate data to present a full patient picture at the point of care, while others like OpenEvidence help ensure that providers make the best decision possible at the point of care. Yet, a lot of these insights are driven by core data that lives within an EHR.
It’s not a large leap to imagine that an EHR could instead own this revenue stream directly, and then orchestrate the administrative and agentic workflows that flow from each clinical decision. The monetization doesn’t have to look like a seat license either. Doximity and OpenEvidence have both shown that alternative business models, including advertising and pharma sponsorship, can work at scale in healthcare software.
Pharmaceutical services
The specialty pharma opportunity may be the most overlooked opportunity by EHRs. Startups and non-EHR incumbents have already built consumption-driven businesses around it, yet most EHR companies have yet to attack this part of the market.
Across high drug-spend specialties (oncology, ophthalmology, rheumatology, dermatology, gastroenterology), significant dollars flow between a physician's prescribing decision and the moment a drug is dispensed. An EHR with full clinical context is positioned to own that entire chain: prior auth, hub enrollment, pharmacy routing, and group purchasing.
Companies like Tandem and Latent have already grown aggressively by automating prior auth, acting as non-dispensing pharmacies, and monetizing through pharmaceutical hub services. To do this, they rely heavily on existing EHR workflows to drive their growth. A new EHR built from scratch has no reason to cede that position. In addition to this workflow, there is also a drug GPO opportunity. Cencora and McKesson have already demonstrated that tying an EHR to a GPO is legally viable, yet no one else has followed. An AI-native EHR that automates the administrative overhead and takes a slice of drug spend would be a fundamentally different kind of business.
How the TAM expands
Not all specialties are created equal for this model of vertically integrating practice management, RCM, clinical decisioning, and pharma workflows.
We believe the specialties most worth targeting have high drug spend per patient, high administrative intensity relative to patient volume and clear clinical AI use cases. The more of those boxes a specialty checks, the more the full vertical OS model makes sense over a pure subscription play. To help bring this to life, we built an illustrative TAM for gastroenterology. Worth flagging that the clinical AI aspect is the most uncertain piece here, given ongoing unknowns around how this care should be delivered and how clinical decisioning will be monetized.
In this framing, pricing the EHR cheaply isn't a bug, it’s a feature that helps drive a more aggressive customer acquisition strategy.
So are ambulatory EHRs a venture opportunity?
When you look at the above TAM, the short answer is yes. The TAMs aren’t massive, but they’re large enough with enough upside to make this category genuinely interesting.
There are still a lot of executional questions to answer around product sequencing. Do you need to build all of this at once? Do you start with a unique wedge? Some combination of a large product build, plus a unique light-weight wedge? There's no universal answer. The answer will heavily be driven by the specialty, the customer, and what you can credibly own from day one.
If you're a provider or operator who sees your practice in any of what we've described above, we want to hear from you. And if you're building or interested in building in this space, we'd love to talk.
- Email: sam@primary.vc & hannah@primary.vc
- LinkedIn: Sam Toole & Hannah McQuaid