More From Green Room
Healthcare AI is attracting substantial investment, with startups building applications for clinical documentation, revenue cycle management, patient scheduling and other administrative workflows.
Clinical documentation was among the first major categories to take off, as AI scribes demonstrated how much administrative work could be automated. Revenue cycle management has since emerged as one of the most active areas of AI adoption, delivering some of the strongest reported financial returns.
Many of these companies initially entered the market as AI layers on top of existing electronic health records (EHRs). They solved a specific problem without requiring healthcare organizations to replace their core software.
But the competitive landscape is evolving. A new generation of AI-native companies is using agentic AI to automate work across multiple underlying systems. Rather than adding a feature to one application, these companies can coordinate tasks across EHRs, billing platforms, payer portals and other software that healthcare organizations rely on every day.
That creates an interesting competitive dynamic. Established EHR providers have the customer relationships, clinical data and distribution to extend AI capabilities within their platforms. Emerging companies, meanwhile, have the opportunity to become a system of action that operates across several legacy systems, without needing to replace them immediately.
Over time, some of those AI-native companies may become the primary software interface for a healthcare organization or replace the EHR entirely.
That raises the investment question we keep coming back to: who will capture the economic value as AI reshapes healthcare software?
We believe established software providers and emerging AI-native companies will both participate. For the latter, the opportunity depends on combining industry expertise, differentiated data, distribution and the ability to execute complex workflows across the systems their customers already use.
The opportunity is not simply to build a better AI feature. It's to become a more valuable operating layer for the healthcare organization.
Healthcare is a large, fragmented vertical market that is adopting AI faster than almost any other, at 2.2 times the rate of the broader economy.[1] US national health expenditure reached $5.3 trillion in 2024, including $1.6 trillion in hospital care and $1.1 trillion in physician and clinical services.[5] AI applications are a much smaller line item: about $1.4 billion in 2025, nearly three times the year before, with health systems accounting for roughly $1 billion of it.[1]
We estimate a rough ceiling for hospital IT budgets of $50 to $80 billion a year, or 3% to 5% of hospital spending. Against that budget, about $1 billion of health system AI spending is under 2%. We believe that gap represents meaningful headroom for AI applications over the next several years.
The opportunity may extend beyond incremental AI spending, however. Companies that become essential to how healthcare organizations operate could eventually capture a larger share of existing software budgets. The size of the market does not tell you which companies will capture that spending.
So far, mostly in the back office. In a 2026 survey of 226 healthcare executives, Bain & Company and Bessemer Venture Partners found buyers reporting a 3.5x return on AI on average, reached in about 12 months against the 24 they had projected.[4] Provider revenue cycle led every category at 4.0x, against 2.9x for provider clinical use cases. Two-thirds of revenue cycle respondents already run semi- or fully autonomous agents. In clinical work, the figure is 4%.[4]
The figures are self-reported, but they line up with the spending data. Revenue cycle and clinical documentation sit at the top of both business priority and AI growth, and together took about $1.05 billion of the $1.4 billion spent on healthcare AI in 2025.[1] Patient engagement is the fastest grower, up about 20x in 2025 from a small base, and we expect patient access to move into the top-right box next. Cybersecurity, on the other hand, is a top priority that AI applications barely touch today.
For investors, the distinction is important. Adoption is accelerating, but the value is not spread evenly. Understanding the underlying workflow, the customer's willingness to pay and the measurable return is still the core of evaluating these businesses.
Revenue cycle is particularly interesting because the work frequently spans multiple systems, from the EHR and practice management platform to payer portals and billing applications. An AI agent that can execute across those systems may offer more value than a feature confined to any one of them.
The most visible use cases are not always where the return is.
Healthcare runs on a long list of administrative workflows: documentation, coding, claims, scheduling, referrals and patient communication. They are attractive entry points for AI companies because the work is manual, the value is easy to measure, and the product can often be added without replacing anything.
Most startups enter the practice as an AI layer on top of the existing EHR. They pick a job where staff time is the bottleneck, such as notes, phones or referrals, and prove ROI without a rip-and-replace. That makes adoption easier.
But the opportunity is increasingly broader than automating one task within one system.
AI-native companies can use agents, including computer-use AI, to work across several underlying applications. A revenue cycle workflow, for example, may require retrieving information from the EHR, checking payer requirements, submitting an authorization and updating billing records. Historically, staff members have moved between these systems to complete the work.
An AI-native layer that can coordinate those steps has the potential to become a system of action across the healthcare organization's existing technology stack. This is a different proposition from adding AI capabilities to an individual software platform.
Incumbents have significant advantages. An established EHR provider already has the customer relationship, the clinical data and integrations across the practice. It can also incorporate AI into its existing product. In August, athenahealth began including its native AI medical scribe in athenaOne at no additional cost.[2]
But a built-in scribe addresses one use case within one platform. It does not, by itself, resolve the broader opportunity to automate complex workflows across multiple systems. The competitive question is whether established providers can extend beyond their own platforms quickly enough to match startups building several agentic applications into a coordinated operating layer.
The data cuts both ways. Bessemer and Bain found established healthcare IT vendors gaining share, most visibly in revenue cycle, yet buyers say systems of record aren't where they expect their next AI to come from. Meanwhile, 57% say they are inundated by AI companies pitching point solutions, and 42% have consolidated or are consolidating vendors.[4]
For an emerging company, the first AI application should be a path to a broader customer relationship, not the whole business. The opportunity is to expand across workflows, coordinate action between systems and become increasingly central to how the customer operates.
Revenue cycle management is a good example of work that is hard to copy. Getting a medical claim paid requires knowledge of payer requirements, procedure codes, authorization rules and denial management, and those rules vary by specialty, payer and provider. Much of that knowledge has historically lived with experienced billing teams, not in the software.
AI that captures that knowledge can improve coding accuracy, speed up claims and reduce administrative cost. Its value comes less from the underlying model than from the data integrations and workflow logic required to run reliably inside a healthcare organization. That is likely why buyers report the strongest returns here.
The ability to operate across multiple systems adds another dimension. These workflows rarely begin and end inside a single application. Successfully automating them requires understanding not only the individual tasks, but how information and decisions move between platforms.
The same principle applies across specialties. An orthopedic practice, a behavioral health provider and a primary care group may buy similar categories of software, but they run quite differently. We believe companies that understand those differences can build products that are harder for general-purpose tools to replicate.
As models become more accessible, specialized data infrastructure, operational expertise and the ability to execute complex workflows across systems become important sources of differentiation. This is consistent with our broader view of AI-native architecture.[8]
The buyer universe is concentrated at the top and fragmented at the bottom. About 397 health systems own 68% of US hospitals and 77% of beds.[6] At the other end, about 42% of physicians work in physician-owned practices, down from 60% in 2012, and 47% still work in practices of 10 or fewer.[7]
For a software company, that means the buyer is increasingly a platform, such as a health system, a private equity-backed group or an MSO, rather than an individual physician. That said, the long tail remains meaningful. Each segment buys differently, and established platforms benefit from distribution, integrations and high switching costs. A new company usually has to earn credibility one specialty or workflow at a time.
For healthcare AI companies, a focused entry point can support expansion into more workflows and a larger share of the customer's spend.
Importantly, that expansion does not necessarily require replacing the systems already in place. An AI-native company can deepen its relationship with a customer by coordinating work across existing platforms, potentially becoming the interface employees use to manage an increasing share of daily operations.
We believe strong retention, customer references and an expanding role in daily operations create a more durable position than superior performance at a single AI task. Bessemer's advice to builders makes the same point: "begin with a single product wedge, earn the workflow, then expand."[4]
Some can, over time. Most begin with one workflow inside or alongside the existing EHR, then expand into scheduling, eligibility, billing and patient communication.
As these applications become more capable, some may become a system of action across multiple legacy platforms. The customer can continue using existing systems of record while relying on the AI-native layer to coordinate and execute work between them. A growing number of companies have an even larger ambition: replacing the EHR outright.
Stage 3 is the bigger prize, as it captures the practice's entire EHR, practice management and billing budget, not just its AI line item. We estimate practices could spend about $9.8 billion a year on software at full adoption, roughly $6.8 billion on SaaS and $2.9 billion, or about 30%, on AI applications.
The hard part is switching costs. Practices rarely change EHRs because of data migration, implementation cost, staff training and operational risk. We expect these companies to win first with new practices, consolidating provider groups and practices whose software is reaching end of life. The best targets tend to be specialties with many independent practices, a fragmented EHR market and heavy administrative work.
The emergence of cross-system AI agents creates an additional path. A company may capture meaningful value by becoming the operating layer across several incumbent platforms before attempting to replace any of them.
It is our opinion that AI-native platforms engineered from scratch will be best positioned to consume portions of the legacy EHR over time. That is consistent with our broader view that AI-native architectures will consume the old SaaS codebase.
Not every successful AI company needs to replace the system of record. But AI-native architecture creates a real opportunity to rethink how core industry software is built, connected and operated.
Healthcare AI has attracted substantial venture investment, particularly in clinical documentation. Competition is increasing, and established software providers are building AI into their own platforms.
These risks put the focus back on the fundamentals: product differentiation, customer retention, capital efficiency and the underlying economics of the business, rather than AI adoption or revenue growth alone.
Healthcare shows many of the characteristics we look for across the essential industries we invest in: government, healthcare, financial services, business services, and construction and infrastructure. These markets run on specialized workflows, fragmented data, legacy software and a lot of human labor. AI creates an opportunity to automate parts of that work and, in some cases, to redesign the software underneath it.
The competitive environment is getting more demanding. Foundation-model providers keep advancing the technology, established platforms are building AI into existing products, and emerging companies are competing across many of the same categories.
But we believe the emergence of AI-native systems of action creates a significant opportunity. Companies that can coordinate work across fragmented legacy environments may capture value that no individual software platform currently addresses.
This puts more weight on the characteristics that have always defined strong vertical software businesses: customer retention, differentiated distribution, deep industry knowledge and an expanding role in the customer's operations.
As we wrote in The Ever-Shrinking Investable Box,[9] growth still counts, but product differentiation, retention, capital efficiency and the quality of the underlying business are increasingly central to investment decisions.
We expect healthcare AI spending to grow significantly over the next several years. The more consequential question is which companies will capture that spending profitably and build valuable customer relationships as AI-native applications evolve into broader operating platforms.
CreativeCo Capital's investment perspective. We believe the strongest emerging vertical AI businesses will combine capable technology with specialized data, industry expertise, distribution and the ability to execute complex workflows across systems.
Some will become indispensable layers on top of existing software. Others may eventually replace portions of the legacy stack entirely.
The battle for the next generation of healthcare software is not simply about who builds the best AI feature. It's about which companies become essential to how healthcare organizations operate.
Travis Parsons is Managing Partner of CreativeCo Capital, a growth equity firm. He writes Market Perspectives for Green Room.
Subscribe to Green Room by CreativeCo Capital for more from our team on private market AI investing.
What is a system of action in healthcare AI?
A system of action is software that can coordinate and execute work across multiple underlying applications, such as EHRs, billing platforms and payer portals. AI-native systems of action can automate workflows that previously required staff to move between disconnected systems.
How big is the healthcare AI market?
Healthcare organizations spent about $1.4 billion on healthcare-specific AI applications in 2025, nearly three times 2024, according to Menlo Ventures. Clinical documentation, at about $600 million, and coding and billing, at about $450 million, were the largest categories.
Why is AI in revenue cycle management attractive for investors?
Revenue cycle management includes coding, claims processing, prior authorization and denial management. These processes are labor-intensive, financially important and frequently span multiple systems. In a 2026 Bessemer Venture Partners and Bain & Company survey, provider revenue cycle led every category with a 4.0x reported return.
Can AI-native companies replace legacy EHR platforms?
We believe some can over time. Many will first build AI layers that operate across existing platforms, allowing customers to automate more work without replacing their core systems. As these companies expand, some may become the primary operating interface or eventually replace portions of the legacy software stack.
How does CreativeCo Capital approach healthcare AI investing?
CreativeCo Capital is a growth equity firm investing in AI-forward companies in essential industries: government, healthcare, financial services, business services, and construction and infrastructure. In healthcare, it evaluates companies on AI-native architecture, industry expertise, customer relationships and the opportunity to expand across complex workflows.
1. Menlo Ventures, 2025: The State of AI in Healthcare, October 2025.
2. HIT Consultant, athenahealth Launches Native AI Ambient Capabilities in athenaOne, August 2026.
3. CreativeCo Capital analysis of disclosed financing rounds for leading healthcare AI documentation companies, October 2026.
4. Bessemer Venture Partners with Bain & Company, The 2026 Healthcare AI ROI Scorecard, September 30, 2026.
5. CMS, National Health Expenditure Highlights, reporting 2024 data.
6. American Hospital Association, Fast Facts on U.S. Health Systems 2025.
7. AMA Physician Practice Benchmark 2024, via Fierce Healthcare.
8. Travis Parsons, What Is AI-Native?, CreativeCo Capital.
9. Travis Parsons, The Ever-Shrinking Investable Box, CreativeCo Capital.
.svg)
