Can healthcare AI agents make money from broke hospitals?

Are US hospitals really too broke to buy healthcare AI?

US hospitals are still struggling financially, but plenty of them can afford healthcare AI agents when the software helps them save or recover more money than it costs.

Kaufman Hall's National Hospital Flash Report, covering more than 1,300 hospitals, reported a 1.3% adjusted operating margin for 2025. That improved to 1.7% through March 2026, although rising bad debt, charity care, and drug expenses continued to put pressure on hospital finances.

At a 1.3% operating margin, a hospital generating $500 million in annual revenue earns just $6.5 million from operations. A $500,000 annual AI contract would absorb nearly 8% of those operating earnings before implementation costs. Even a relatively small software purchase can therefore become a serious financial decision.

The situation is worse among smaller providers. The American Hospital Association's 2026 Environmental Scan reported that 48% of rural hospitals operated at a financial loss in 2023. Research from the Center for Healthcare Quality and Payment Reform also found that approximately one-third of rural hospitals were at risk of closing.

But those figures hide enormous differences between hospital groups. A large academic medical center with several billion dollars in revenue has considerably more room to invest than a rural hospital operating on a few million dollars of annual income.

For healthcare AI startups, the biggest commercial opportunity lies with health systems that handle enough administrative work to generate substantial savings. The hospitals under the greatest financial pressure may need automation most, but they are often the hardest customers to sell to.

Why are hospitals buying AI agents when they are cutting costs?

Hospitals are investing in AI agents because reducing administrative expenses has become one of the few ways to improve their finances without cutting essential medical services.

According to the American Hospital Association, workforce expenses represented 56% of hospital costs in 2024. Hospitals can reduce some administrative workloads through automation, while cutting clinical staff is much harder without affecting patient care.

We can already see this in purchasing intentions. Oliver Wyman's 2026 healthcare revenue cycle survey, covering more than 200 decision-makers and 90 end users, found that 70% to 90% of decision-makers expected to increase spending on AI-enabled revenue cycle capabilities over the following three years.

That's a striking result for an industry facing persistent financial pressure. Some of that money will come from budgets previously allocated to outsourced billing, temporary administrative workers, and older software.

However, those survey results measure planned spending rather than completed purchases. We still need to see how much of that interest translates into signed contracts, larger deployments, and recurring vendor revenue.

What jobs are healthcare AI agents actually replacing?

Healthcare AI agents are taking over parts of medical documentation, insurance administration, billing, and patient communications, with administrative work moving toward automation faster than clinical decision-making.

The difference between an AI assistant and an AI agent matters here. An ambient scribe that listens to a consultation and drafts a note saves time, but the clinician generally remains responsible for reviewing and signing it. An agent that checks insurance eligibility, retrieves missing information, and updates a billing workflow can complete a larger share of the task.

Abridge has built its business around clinical documentation and increasingly connected clinical workflows. AKASA focuses on revenue cycle processes, especially medical coding and documentation accuracy. Infinitus automates healthcare communications, including repetitive calls involving insurers, providers, and patients.

These companies serve different needs, and the savings they promise are different too.

Documentation software may allow doctors to spend less time writing notes or completing records. Revenue cycle agents can reduce billing expenses and improve collections. Communications agents can replace repetitive calls and administrative follow-ups.

The important question is how much of the original human work each product actually eliminates. A system that prepares information for an employee may be useful, but its financial value is usually lower than one that completes the task without additional intervention.

Are hospitals actually using AI agents at scale, or just testing them?

Healthcare AI has moved beyond pilots in clinical documentation and some administrative workflows, although fully autonomous hospital operations are still far from widespread.

Oliver Wyman's 2026 survey found that approximately 20% to 40% of provider organizations reported broad or enterprise-wide adoption of AI-enabled tools across different revenue cycle activities. Adoption varied considerably by workflow, and many hospitals were still using automation within individual departments.

Several vendor deployments support this picture. Abridge reported more than 150 enterprise health system deployments in its ambient AI impact report. Ambience Healthcare also announced an enterprise rollout with MultiCare Health System after an evaluation involving more than 550 clinicians across over 20 specialties.

MultiCare's decision is particularly interesting because the hospital system compared three ambient AI products during 2025. The evaluation went beyond checking whether AI could produce acceptable documentation. It considered which product performed better in actual clinical work.

A hospital using AI documentation across thousands of clinicians has achieved considerable adoption. But that doesn't mean its billing department, scheduling team, or insurance authorization processes are operating autonomously.

Which healthcare AI agents can make hospitals the most money?

Revenue cycle AI agents currently have the strongest financial case because hospitals can measure their impact through recovered payments, fewer billing errors, and lower administrative costs.

Experian Health's 2025 State of Claims survey found that 41% of surveyed healthcare providers had claim denial rates of at least 10%, compared with 30% in 2022. The same survey found that 43% were struggling with staffing shortages and that 90% of denied claims received some human review before resubmission.

That combination makes insurance administration an obvious target. Hospitals face a growing number of rejected claims while relying on employees to investigate and correct them.

Medical coding offers another opportunity because incorrect or incomplete documentation can affect legitimate reimbursement. AKASA has developed AI systems specifically for complex inpatient coding, where accuracy and compliance directly affect hospital finances.

Clinical documentation can also produce financial returns. A study involving St. Luke's Health System and Ambience Healthcare reported approximately $13,000 in annual coding-related financial benefit per clinician, connecting better documentation with reimbursement outcomes.

Patient communications and scheduling are further opportunities, although their financial results depend more heavily on whether automation replaces paid work or increases completed appointments.

AI agent application	Main financial opportunity	Evidence or constraint
Claim denial prevention	Protect collectible revenue	41% of surveyed providers report denial rates of at least 10%
Medical coding	Improve coding accuracy and reduce manual work	AKASA reports autonomous coding for complex inpatient records
Clinical documentation	Save clinician time and support accurate reimbursement	St. Luke's reported $13,000 in annual coding-related benefit per clinician
Prior authorization	Reduce administrative effort and delays	35% of surveyed providers cited authorization problems as a denial cause
Insurance eligibility checks	Prevent billing errors before treatment	32% cited inaccurate or incomplete registration data
Patient scheduling	Reduce administrative costs and unused appointment capacity	Returns depend on completed visits and actual labor savings
Clinical decision support	Improve clinical work and decision quality	Direct cash savings remain harder to measure

If you want more recent data on this point, please see our latest healthcare AI agents market report.

Can AI agents really recover millions in lost hospital revenue?

Hospital revenue cycle AI agents can potentially recover millions of dollars annually, even with small improvements in collections, but vendors must prove that the money would otherwise have been lost.

Experian Health's 2025 survey identified authorization problems, registration errors, and missing or inaccurate information as major contributors to claim denials.

These problems are well suited to automation because many involve checking information against existing records and insurance requirements.

Consider a hospital generating $500 million in annual net patient revenue. If an AI system increases actual net collections by just 0.2%, the hospital receives an additional $1 million per year. A 0.5% improvement produces $2.5 million.

Those are illustrative calculations, rather than observed average results from AI deployments. They demonstrate the size of the opportunity.

Using Kaufman Hall's 1.3% operating margin as a reference, a hospital with $500 million in revenue would earn approximately $6.5 million from operations. An extra $1 million in net operating results, assuming no offsetting costs, would increase those earnings by about 15%.

But there's a catch with claims recovery.

Many rejected claims eventually get paid after employees correct and resubmit them. An AI system that prevents those denials may save labor and accelerate payments, but counting the entire value of those claims as new revenue would exaggerate its contribution.

We should therefore judge revenue cycle agents on verified additional net collections, lower write-offs, reduced administrative costs, and faster payments. Gross claim values alone tell us very little about how much money the hospital actually makes.

Are AI medical scribes saving hospitals money or just making doctors happier?

AI medical scribes are already reducing clinician burnout and documentation work, but their direct financial returns are much harder to establish than those of billing automation.

A multicenter study published in JAMA Network Open in 2025 examined 263 clinicians across six healthcare systems. After 30 days of using ambient AI scribes, the proportion reporting burnout fell from 51.9% to 38.8%, a reduction of 13.1 percentage points.

That's a meaningful improvement for hospitals struggling to retain clinicians. However, lower burnout does not automatically translate into lower spending.

A randomized crossover trial involving 160 outpatient clinicians found differences between two ambient scribe products in workflow satisfaction and documentation time, but no meaningful differences in after-hours electronic health record time.

The results show why the financial case varies across products and deployments. Saving a physician 20 minutes of documentation may improve working conditions without reducing payroll costs or creating additional appointments.

The picture changes when documentation software improves billing accuracy.

A KLAS Research study involving St. Luke's Health System and Ambience Healthcare reported a 41% reduction in chart closure time alongside approximately $13,000 in annual coding-related financial benefit per clinician.

If 500 clinicians achieved the same financial benefit, the total would reach $6.5 million annually. That calculation is illustrative, and actual results would depend on clinician mix, reimbursement, and deployment costs.

Are healthcare AI agents actually cheap enough to be profitable?

Healthcare AI agents can be profitable when they automate large volumes of repetitive work, but hospitals and startups need to account for integration, supervision, and difficult cases alongside AI processing costs.

A hospital pays for completed work. The cost of generating an AI response is only one part of delivering that work.

An agent may need to retrieve patient information, access insurance systems, interpret policy requirements, update hospital records, and escalate complicated cases to employees.

If the system completes 90% of tasks independently, the remaining 10% may be considerably more expensive than average because the difficult cases require experienced staff.

We can model the potential economics of a hospital processing one million administrative tasks annually, each previously costing $4 in labor.

Operating scenario	Annual amount
Existing administrative labor cost	$4,000,000
Tasks completed independently by AI at 80% automation	800,000
Gross avoided labor value at $4 per task	$3,200,000
Remaining manual workload	$800,000
AI infrastructure and usage costs, assumed	$300,000
Integration, supervision, and support, assumed	$400,000
Potential savings before vendor subscription	$2,500,000

These figures are hypothetical and assume that the hospital can actually eliminate the equivalent labor expenses. If employees simply spend the saved time on other work, the accounting savings may be much smaller.

Under these assumptions, a $1 million annual vendor fee would leave the hospital with $1.5 million in potential savings.

For the startup, however, that same $1 million is revenue rather than profit. The vendor still needs to cover its own technology costs, implementation teams, customer support, and product development.

If you want more recent data on this point, please see our latest healthcare AI agents market report.

Can hospitals benefit from AI agents without firing employees?

Healthcare AI agents can deliver financial value without layoffs by reducing overtime, avoiding future hires, and allowing hospitals to handle more work with existing staff.

Many hospitals are dealing with staffing shortages rather than having excess employees. Repetitive administrative work can occupy people who could otherwise focus on more complicated cases.

Infinitus provides an example through healthcare communications. The company reported that its AI platform automated calls on behalf of 1.17 million patients in 2025.

That level of activity suggests AI can handle significant volumes of repetitive coordination work. But the number of automated calls does not reveal how much hospitals or other healthcare customers actually saved.

If the calls replace outsourced services or reduce overtime, the savings can be direct. If they simply allow an existing team to process more cases, the financial benefits depend on whether the additional capacity increases revenue or avoids future spending.

For hospital buyers, avoiding five planned administrative hires can be a valid return even if no existing employee loses a job.

Can small rural hospitals afford healthcare AI agents?

Rural hospitals are unlikely to become major direct customers for expensive healthcare AI agents unless vendors can offer lower-cost products with minimal implementation work.

The American Hospital Association has repeatedly documented financial pressure among rural providers. Its 2026 Environmental Scan highlighted the high proportion of rural institutions operating at a loss, while the Center for Healthcare Quality and Payment Reform warned that many remained at risk of closing.

Patient volumes are also much smaller than at major health systems.

A $200,000 annual AI contract might make sense for a large organization processing hundreds of thousands of claims. For a rural hospital handling a fraction of that volume, the same price could be impossible to justify.

Still, smaller hospitals could benefit from administrative automation, particularly when they have difficulty recruiting specialized billing or coding staff.

The better commercial approach may be to distribute AI through regional hospital networks, outsourced revenue cycle companies, or existing software providers. A single technology platform could serve dozens of smaller hospitals without requiring separate engineering teams for each installation.

If you want more recent data on this point, please see our latest healthcare AI agents market report.

Are healthcare AI startups making real money yet?

Healthcare AI startups are winning substantial enterprise contracts and deployments, but publicly available information still doesn't establish that most AI-native vendors are profitable.

Abridge is one of the strongest examples of commercial adoption. The company reported more than 150 enterprise health system deployments, while CB Insights estimated its 2025 revenue at approximately $117 million.

That revenue estimate comes from a third party rather than an audited company disclosure, so it should be treated cautiously. Still, combined with Abridge's deployment figures, it suggests a business that has progressed well beyond small experiments.

Ambience Healthcare raised $243 million in a Series C round in July 2025. Its subsequent selection by MultiCare following a competitive evaluation provided further evidence that hospitals are purchasing specialized clinical AI products.

Yet neither a large financing round nor a major customer announcement proves profitability.

For a stronger financial reference, we can examine Waystar, an established healthcare payments software company rather than an AI-native startup.

Waystar reported $1.099 billion in revenue for 2025, approximately $112 million in net income, and $462 million in adjusted EBITDA. Its adjusted EBITDA margin was roughly 42%.

Those figures establish that healthcare software businesses can produce substantial profits while serving hospitals and other providers facing financial pressure.

There is still a major difference between Waystar's mature platform and a venture-backed startup building and deploying AI agents. The startup may have considerably higher development costs relative to revenue, less predictable contracts, and more expensive implementations.

Could Epic, Microsoft, and Waystar crush healthcare AI startups?

Epic, Microsoft, and Waystar are serious threats to independent healthcare AI startups because they already have access to hospital workflows, large customer bases, and established distribution channels.

Waystar offers a particularly useful comparison. Its 2026 agentic AI announcement described a network handling more than 7.5 billion annual transactions and touching approximately one in three US hospital discharges.

The company also reported that its AltitudeAI technology had helped prevent $15.5 billion in denials in less than one year. That is a vendor-reported measure of prevented denials, rather than verified additional cash collected, but it shows the scale at which the platform operates.

Waystar's financial results reinforce its advantage. In the first half of 2026, the company reported $633.5 million in revenue, up 20.2% year over year. It also disclosed 108.3% net revenue retention and more than 1,400 customers generating at least $100,000 in trailing annual revenue.

An AI startup entering the same market must spend time securing data access, meeting security requirements, integrating systems, and proving it can be trusted.

Waystar can develop AI capabilities inside an existing payments network. Epic benefits from its position inside hospital electronic health records, while Microsoft can distribute clinical AI through products and commercial relationships that hospitals already use.

That advantage becomes even greater when established vendors package AI into broader software contracts.

Specialized startups still have room to win. A product that handles complex medical coding, specialty documentation, or insurance workflows considerably better than an incumbent can justify a separate contract.

If you want more recent data on this point, please see our latest healthcare AI agents market report.

Can healthcare AI agents connect to hospital software without expensive custom work?

Connecting healthcare AI agents to existing hospital systems is one of the biggest barriers to scaling, especially for startups that must build custom integrations for each customer.

A hospital agent may need access to patient records, billing applications, insurance portals, appointment systems, and internal approval processes.

These systems handle sensitive information and were often designed around employees following established workflows. Autonomous software needs carefully controlled access, audit trails, and reliable ways to manage errors.

The leading vendors have adapted their products accordingly. Abridge emphasizes integration with clinical documentation workflows inside electronic health records. AKASA develops technology around specialized coding and revenue cycle processes. Waystar can use the infrastructure of its existing payments network.

An independent startup without those advantages may need substantial engineering and support work for every deployment.

That can seriously damage margins. A company may sell an expensive annual subscription but spend much of the contract's value on implementation, integration, and ongoing support.

Clinical accuracy creates another cost. AKASA has reported third-party evaluations in which its autonomous inpatient coding technology matched or exceeded expert coder accuracy, while estimating that 65% to 85% of inpatient volume at most health systems could be addressed autonomously.

Those results suggest a significant opportunity, although the remaining cases still require appropriate human involvement.

Healthcare security adds further complexity. An Imprivata survey reported in 2026 that 72% of healthcare industry leaders had encountered AI tool use without formal IT approval, illustrating the governance problems hospitals must manage.

If connecting the twentieth hospital costs almost as much as connecting the second, the business may struggle to achieve the margins normally associated with enterprise software.

Will hospitals pay healthcare AI agents monthly or only when they deliver results?

Healthcare AI pricing is moving toward a combination of subscriptions, usage fees, and performance-linked contracts, because different hospital workflows produce different kinds of measurable value.

Clinical documentation is naturally suited to per-clinician pricing. Hospitals know how many doctors use the software and can compare subscription expenses with time saved or other documented benefits.

Claims processing and insurance verification are better suited to transaction-based fees because the number of completed tasks is measurable.

Revenue recovery creates an opportunity for performance-linked contracts, where the vendor receives a share of verified additional collections or an agreed amount for successful outcomes.

That approach is attractive to financially constrained hospitals because it reduces the risk of paying for technology that delivers disappointing results.

However, performance pricing is complicated. A hospital and vendor must agree on what would have happened without AI, which recovered payments are genuinely incremental, and how to account for existing staff contributions.

We can compare the main approaches.

Pricing model	Best application	Main challenge
Per-clinician subscription	Clinical documentation	Usage and financial value vary by specialty
Enterprise subscription	Broad hospital AI platform	Requires a large budget commitment
Per completed transaction	Claims processing and insurance verification	Revenue depends on completed volume
Per automated call	Patient and insurance communications	Calls do not always replace paid work
Share of recovered revenue	Claim denial and underpayment recovery	Difficult to measure incremental collections
Share of verified savings	Administrative cost reduction	Both parties need an agreed baseline
Subscription plus usage fees	Multiple administrative workflows	Hospitals need predictable spending

Established healthcare technology companies already demonstrate the appeal of recurring and transaction-based revenue.

Waystar's 2026 regulatory filing reported that more than 99% of its revenue was recurring subscriptions or came from highly predictable transaction volumes.

For AI-native startups, a minimum subscription combined with fees for completed work may be particularly attractive. It gives the vendor predictable income while allowing hospitals to connect part of their spending to actual activity.

Performance-based pricing could become more common as vendors accumulate reliable data on outcomes, but it's unlikely to replace subscriptions across every healthcare AI category.

Can healthcare AI startups survive long hospital sales cycles as prices fall?

Healthcare AI startups can survive lengthy hospital evaluations and growing price competition, but they need deployments to become cheaper and customer adoption to expand faster than prices decline.

MultiCare's selection of Ambience Healthcare followed a head-to-head evaluation involving more than 550 clinicians across over 20 specialties. Such evaluations can support major contracts, but they also require considerable time and resources before a startup earns recurring revenue.

Abridge's large number of enterprise deployments shows that hospitals can repeatedly adopt the same technology. What remains less visible is how much implementation work each new customer requires and how quickly signed contracts turn into revenue.

Meanwhile, ambient documentation is becoming increasingly competitive. Abridge, Ambience Healthcare, Microsoft's Dragon Copilot, Nabla, and other vendors now offer overlapping capabilities.

MultiCare's comparison of three solutions shows how hospitals can use that competition to evaluate quality, negotiate prices, and demand more functionality.

Lower prices are not necessarily disastrous. A vendor could charge less per clinician while increasing revenue through broader deployment or additional workflows. But startups selling features already included in existing hospital software contracts will have a harder time defending premium pricing.

Complex processes involving payer rules, medical coding, clinical data, and exception handling may offer stronger protection from commoditization than basic documentation tools.

The commercial test is whether an AI startup can expand inside existing hospital customers while reducing implementation costs and retaining enough pricing power to build healthy margins.

Can healthcare AI agents make money from broke hospitals?

Yes. Healthcare AI agents can build profitable businesses selling to financially constrained hospitals, particularly when they automate billing, coding, insurance administration, and other work that directly affects costs or revenue.

The evidence is stronger than it was during the early wave of hospital AI experiments.

As we saw above, Kaufman Hall documents thin hospital operating margins, while Experian Health shows that claim denials and administrative problems remain expensive. Yet healthcare providers are still planning higher AI spending, and vendors such as Abridge and Ambience Healthcare have secured substantial enterprise deployments.

The most attractive business models are emerging around measurable financial outcomes. A hospital with hundreds of millions of dollars in revenue can justify a substantial AI contract if the technology produces even modest improvements in legitimate collections or administrative expenses.

Clinical documentation is also a real market, but vendors relying primarily on time savings and clinician satisfaction may face greater pressure on pricing. Products connecting documentation to accurate coding, reimbursement, and productivity have a stronger argument.

The biggest unresolved question is startup profitability. Waystar demonstrates that healthcare payments software can generate billion-dollar revenue and strong operating margins. AI-native companies have shown growing commercial adoption, but most have yet to disclose the financial results needed to prove comparable profitability.

Competition will also get harder. Epic, Microsoft, Waystar, and other established vendors already have the hospital relationships and integrations that independent startups must build. Generic AI capabilities will become increasingly difficult to sell at premium prices.

Our judgment is that healthcare AI agents have a genuine path to profitability, but the opportunity is narrower than the excitement around hospital automation suggests. The companies most likely to succeed will sell measurable improvements in hospital finances, deploy across customers without expensive custom work, and retain enough of the savings they create to support healthy margins.

Hospitals may be short of money, but they already spend enormous amounts on administration. Healthcare AI startups can make money by taking over part of that work more cheaply. The harder challenge is proving that they can do it consistently, at scale, and with enough profit left over for themselves.

If you want more recent data on this point, please see our latest healthcare AI agents market report.