Is % of collections the best AI billing business model?

Why are AI medical billing startups so interested in percentage-of-collections pricing?

Percentage-of-collections pricing is especially attractive for AI medical billing startups because it lets them earn a share of healthcare payments while using automation to reduce the cost of processing those payments.

A medical practice collecting $10 million annually generates $500,000 in vendor revenue at a 5% fee. That's an unusually large contract for a software company serving a single healthcare customer, particularly if much of the billing work can be automated.

The model existed long before generative AI. Traditional revenue cycle management (RCM) companies such as R1 RCM and athenahealth built substantial businesses around collecting an agreed percentage of providers' revenue. AI companies are now going after the labor-intensive work that made those fees necessary in the first place.

There's one complication: traditional RCM contracts often cover almost everything between treating a patient and receiving payment. An AI startup that only automates claim submission is selling something much narrower. Whether percentage pricing is attractive depends heavily on how much of that work the vendor actually takes over.

What are healthcare providers actually paying for with a percentage-of-collections contract?

A percentage-of-collections contract can cover the entire medical billing operation or just selected services, and that difference can completely change whether the price is reasonable.

R1 RCM's 2023 annual report described a model in which some operating fees were calculated using an agreed cost-to-collect percentage applied to defined patient collections, alongside contractual adjustments and operating expense arrangements.

A full-service contract may include medical coding, claim submission, denial management, payment posting, patient billing, collections, and the employees needed to perform those tasks. A narrower AI product might handle only two or three of them.

Even the definition of collections varies. Contracts may include insurance payments, patient payments, or revenue from specific facilities. Refunds, historical receivables, and some reimbursement categories can be excluded.

For a practice collecting $10 million, a 5% fee means $500,000 annually. That's potentially reasonable for replacing a billing department and much harder to justify for automating a single administrative task.

Is charging 5% of collections more profitable than selling a subscription?

A 5% collections fee can generate far more revenue per healthcare customer than a software subscription, although higher revenue doesn't necessarily mean higher profit.

Tebra, for example, has publicly marketed monthly subscription packages for medical practices. Its software model offers a very different way of monetizing billing and practice-management technology compared with traditional percentage-based outsourcing.

Consider a hypothetical practice collecting $5 million annually. We can calculate what several possible contracts would generate for the vendor. These are illustrative pricing scenarios rather than actual offers from the companies mentioned.

Billing model	Pricing assumption	Annual vendor revenue
Collections-based	3% of collections	$150,000
Collections-based	4% of collections	$200,000
Collections-based	5% of collections	$250,000
Collections-based	6% of collections	$300,000
Fixed subscription	$799/month	$9,588
Fixed subscription	$2,500/month	$30,000
Per-claim	$2 on 20,000 claims	$40,000
Per-claim	$5 on 20,000 claims	$100,000
Hybrid	$5,000/month + 1%	$110,000

The difference is enormous. A 5% fee generates more than eight times the revenue of a $2,500 monthly subscription in this example.

But the comparison only works as a revenue illustration. A full-service billing vendor might need employees, compliance specialists, integrations, and continuous customer support, while a subscription provider could be selling software that the practice operates itself.

If you want more recent data on this point, please see our latest AI revenue cycle management market report.

Will medical practices really pay AI billing companies 3% to 7% of collections?

Medical practices already pay several percentage points of revenue for outsourced billing, so AI providers can sell within an established pricing market. Getting those same fees for software alone is much less certain.

Traditional billing providers frequently advertise percentage-based rates, with the actual fee varying by specialty, claims volume, complexity, and the responsibilities being outsourced. Higher rates often reflect more difficult billing work or smaller practices with less volume.

An important reference is athenahealth's published estimate that medical practices typically spend approximately 3% to 4% of their revenue on collections-related costs. That isn't a universal industry rate, but it helps explain the economics that vendors must compete against.

A practice currently spending 4% on billing has a strong reason to consider an automated service charging 3% if the service can handle comparable work. Paying an additional 3% for AI software on top of its existing billing expenses is a different proposition.

We're therefore looking at two separate markets: AI companies replacing existing RCM spending and software vendors trying to sell additional tools to organizations that still operate their own billing departments.

Can AI make percentage-of-collections billing dramatically more profitable?

AI can make collections-based billing substantially more profitable by cutting manual processing costs, and existing administrative automation benchmarks show just how large those savings can be.

The 2024 CAQH Index estimates that processing a claim-status inquiry manually costs $13.80, compared with $3.64 electronically. Eligibility verification costs fall from $8.57 to $2.00, while claim submission drops from $6.33 to $3.05.

Those represent estimated savings of 74%, 77%, and 52%, respectively. They measure electronic administrative transactions rather than the incremental effect of generative AI, but they establish how expensive manual processes are.

The potential margin improvement becomes clearer with a simple calculation. An AI billing vendor generating $500,000 annually from one customer would earn a gross margin of 80% if its direct delivery costs were $100,000. If the vendor still needed $300,000 in labor, infrastructure, and support, its gross margin would fall to 40%.

Traditional RCM providers often carry substantial labor costs. AI-native companies may eventually deliver comparable services with much smaller teams, although we still need better financial disclosures to know how often that is happening.

Are AI billing companies actually replacing human billing teams yet?

AI medical billing companies are already automating substantial amounts of work, but the publicly available evidence doesn't yet show widespread elimination of entire revenue-cycle teams.

Candid Health has advertised a touchless claim rate above 95%, meaning that, under its definition, more than 95% of claims were submitted and finalized without human intervention. The company has also reported a 97% payer net collection rate. Both are company-reported performance figures.

AKASA provides a different example. Its case study with Methodist Health System describes more than 56,000 accounts processed over eight months, over 5,000 hours of work saved, and 71.3% of accounts removed from staff queues. The deployment covered a health system with four hospitals and 33 clinics.

These results are impressive but measure different activities. Candid's claim metric concerns a defined claims workflow, while AKASA's study focuses on automating account-status work.

Billing still includes difficult appeals, coding exceptions, payer disputes, patient collections, and payment reconciliation. Companies are making real progress on specific workflows, but a high touchless-claim percentage doesn't mean an entire billing operation runs without people.

Is the healthcare billing market big enough to make percentage-based AI companies huge?

The US healthcare billing market is easily large enough to support multibillion-dollar AI companies, although the amount AI vendors can realistically capture is much smaller than total healthcare spending.

According to CMS, American hospitals accounted for approximately $1.635 trillion in healthcare spending in 2024, while physician and clinical services represented another $1.110 trillion. Combined, that's about $2.74 trillion.

If vendors charged 1% against an equivalent payment base, the resulting annual fee pool would reach approximately $27.4 billion. At 3%, it would exceed $82 billion.

These calculations illustrate the scale rather than measure an addressable market. Healthcare expenditures aren't identical to eligible collections, and hospitals already use internal teams, software platforms, and outsourcing contracts.

Waystar offers a more tangible measure of the commercial opportunity. The healthcare payments software company generated $1.099 billion in revenue in 2025, up 16.5% from the previous year, without relying on a universal percentage-of-collections fee.

We don't need to assume that AI will capture a large fraction of medical payments for this to become an enormous software category. Replacing even part of today's administrative spending could support several large businesses.

Can AI billing companies earn more money without charging a percentage of collections?

Subscription and transaction-based billing companies can generate highly recurring revenue, and Waystar's recent financial results offer strong evidence.

According to its 2025 annual report, more than 99% of Waystar's revenue came from subscriptions or highly predictable transaction volumes. Its net revenue retention reached 112%, meaning existing customers, in aggregate, were spending 12% more after accounting for expansion, contraction, and departures.

Waystar generated $558.4 million in subscription revenue and $534.8 million in volume-based revenue during 2025. Subscriptions grew 22% from the previous year, compared with 11% for volume-based services.

Subscription revenue was growing roughly twice as fast, even though both models benefited from the same broader demand for healthcare payment technology.

The company also reported 1,391 customers generating more than $100,000 each in annual revenue, up 16% year over year.

These figures make it difficult to argue that collections-based commissions are necessary for recurring healthcare billing revenue. Customers will keep paying for valuable infrastructure, regardless of whether the invoice is tied directly to their medical collections.

Do AI billing vendors automatically benefit when their customers collect more money?

Percentage-based AI billing companies earn more when healthcare providers collect more, even when the additional payments have little to do with better billing performance.

Suppose a practice collects $10 million annually and pays a 5% fee. The billing vendor earns $500,000. If collections rise 10%, vendor revenue climbs to $550,000.

That $50,000 increase could reflect better denial management, but it could just as easily come from more patient visits, higher reimbursement rates, or the acquisition of another practice.

This creates an incentive problem. The vendor receives more money even when the healthcare provider generated the growth independently.

The same issue becomes more pronounced with high-value procedures. Two claims may require similar processing work while producing very different payments. A percentage fee automatically charges more for the larger claim.

For companies taking responsibility for the entire revenue cycle, that arrangement may still make sense. For vendors selling narrowly defined automation, it's a weakness that customers can eventually challenge.

If you want more recent data on this point, please see our latest AI revenue cycle management market report.

Could percentage-of-collections contracts become too expensive as medical practices grow?

Percentage-of-collections pricing can become a bad deal for successful healthcare practices when vendor fees rise much faster than the actual work required.

Imagine a practice growing from $5 million to $20 million in annual collections. At a fixed 5% rate, the billing bill rises from $250,000 to $1 million.

If the practice also processes four times as many claims, the vendor has more work to do. But if much of the growth comes from higher reimbursement per procedure or a change in treatment mix, the workload might increase far less than the fee.

The practice now has a strong incentive to renegotiate, demand volume discounts, or consider software charged by provider, claim, or transaction.

There's another practical consequence: large healthcare organizations usually have more resources to compare vendors and negotiate contracts. They may be less willing than small practices to accept an unchanged percentage as their revenue grows.

Percentage pricing can be especially attractive during the early years of a customer relationship, but long-term revenue per customer may not keep rising at the same rate as medical collections.

Is charging a percentage of recovered revenue better than taking a cut of everything?

For AI companies specializing in denial recovery or underpaid claims, charging a percentage of additional money recovered is often easier to justify than taking a share of all collections.

Consider a healthcare provider with $10 million in eligible receivables that historically collects $9.5 million. An AI system helps raise collections to $9.8 million, generating an additional $300,000.

A 20% performance fee on that improvement produces $60,000 in vendor revenue. By contrast, a 5% fee on all $9.8 million collected would generate $490,000.

The first contract has a clearer link to the financial improvement delivered. That's particularly useful for vendors handling old unpaid claims, appeals, and underpayment recovery.

The difficult part is proving how much of the improvement came from the AI system. Payer reimbursement changes, claim age, case mix, and payment timing can all influence results.

A performance contract therefore needs a well-defined baseline, eligible claim categories, and rules for attributing recovered payments. Without those controls, providers and vendors can end up disagreeing about what the software actually earned.

Could per-claim pricing beat percentage-of-collections billing?

Per-claim pricing is a serious threat to percentage-based AI billing because claim-processing costs are often more closely related to transaction volume than payment size.

A system handling two claims with similar documentation and payer rules may perform roughly the same amount of work even if one claim pays $200 and another pays $2,000.

At a 5% collections fee, the vendor receives $10 for the first claim and $100 for the second. At $2 per processed claim, it receives $2 for either.

The CAQH administrative benchmarks reinforce the importance of transaction economics. Electronic claim submission costs an estimated $3.05 per transaction, compared with $6.33 manually. These figures aren't commercial prices for AI billing software, but they show why standardized processing can become much cheaper.

As automation improves, routine claim submission and eligibility verification are likely to face stronger price competition. Providers can compare the cost of individual transactions across suppliers without needing to evaluate an entire outsourced billing operation.

Percentage pricing should remain harder to displace for vendors managing complex collections, appeals, and full operational responsibility.

If you want more recent data on this point, please see our latest AI revenue cycle management market report.

Can subscription-based AI billing software be more profitable than percentage-based RCM?

Subscription-based healthcare billing software can be highly profitable, and Waystar shows that selling recurring software and transaction services can produce strong margins at scale.

In its February 2026 financial results, Waystar reported $1.099 billion in annual revenue, $462.1 million in adjusted EBITDA, and $112.1 million in net income. Its adjusted EBITDA margin reached 42%.

Revenue had grown from $791 million in 2023 to $943.5 million in 2024 and $1.099 billion in 2025, an increase of approximately 39% over two years.

The company also generated $310 million in operating cash flow during 2025. Its financial performance demonstrates that healthcare payment software can combine recurring revenue with substantial operating profits.

Waystar's profitability reflects both subscription and volume-based services, and the company doesn't separately report the margin earned by each pricing model. Its results also don't establish that AI-native startups will achieve comparable margins.

Still, a billion-dollar company with a 42% adjusted EBITDA margin provides strong evidence that taking a percentage of medical collections isn't the only attractive way to monetize healthcare billing.

Can AI billing startups undercut traditional RCM companies and still make money?

AI-native billing companies have a credible opportunity to charge less than traditional RCM providers while earning attractive margins, provided they automate enough work to avoid rebuilding the same labor-heavy operation.

Traditional providers such as R1 RCM have handled large portions of customers' revenue-cycle operations, sometimes taking responsibility for personnel and third-party services. These contracts are difficult to replace because the provider manages workflows that directly affect healthcare organizations' cash flow.

AI companies are approaching the problem differently. Candid Health focuses on automating claims and billing operations, while AKASA has demonstrated productivity gains in administrative revenue-cycle work. Waystar is embedding AI into its established healthcare payments infrastructure.

Their products and responsibilities differ, but together they illustrate three routes into the market: replacing complete billing operations, automating selected tasks, or selling AI through an existing platform.

The clearest opportunity for an AI-native provider is replacing a traditional 5% billing arrangement with a cheaper service that delivers comparable collections.

A startup charging 3% on $10 million in annual collections would still generate $300,000 from that customer while potentially saving the healthcare provider $200,000 annually. Whether the startup can make that contract profitable depends on its actual delivery costs.

Will AI eventually force medical billing fees below 3%?

AI should put growing pressure on routine medical billing fees, but there isn't enough evidence to claim that the industry is already moving toward a universal rate below 3%.

The cost reductions seen in electronic administrative transactions show that significant savings are possible. More recent AI deployments suggest that additional manual work can be removed from claims processing and account management.

However, billing vendors still carry integration, compliance, customer support, and exception-handling costs. Healthcare providers also face meaningful disruption when switching systems, particularly if a vendor is responsible for collecting millions of dollars in payments.

The competitive pressure will probably appear unevenly. Standardized tasks such as eligibility checks, claim submission, and payment posting are easier to compare and price per transaction. Difficult appeals, unusual coding cases, and complex payer disputes remain harder to automate.

Waystar's 2025 growth also shows that established healthcare payment platforms continue to attract customer spending even as AI improves.

We expect the largest fee reductions where AI makes processing predictable and highly automated. Full-service RCM pricing should move more slowly because the vendor still assumes responsibility for the broader billing operation.

If you want more recent data on this point, please see our latest AI revenue cycle management market report.

Which AI billing pricing model makes the most sense today?

The best AI billing pricing model depends primarily on whether the vendor is selling software, completing individual transactions, recovering unpaid revenue, or taking over the customer's entire billing operation.

Our comparison favors collections-based pricing for full-service RCM, per-claim pricing for standardized workflows, and performance fees for clearly measurable recoveries. For broader software platforms, subscriptions and hybrid contracts offer more flexibility.

Billing model	Biggest advantage	Main problem	Best suited for
% of total collections	High recurring contract value	Fees can outgrow workload	Full-service RCM
% of incremental collections	Payment tied to improvement	Hard to measure attribution	Revenue optimization
Flat subscription	Predictable recurring revenue	Heavy users may cost more	Billing software
Per claim	Tracks processing volume	Price competition	Claims automation
Per transaction	Clear unit economics	Lower revenue per task	Eligibility and payments
Per provider	Simple budgeting	Weak link to usage	Small practices
Per recovered dollar	Clear customer ROI	Recovery disputes	Denial management
Hybrid base + variable fee	Combines predictable and variable revenue	More complex contracts	Integrated AI RCM

Is percentage of collections the best AI billing business model?

Partly true. Percentage of collections is one of the most attractive business models for AI companies replacing complete medical billing operations, but the evidence favors hybrid pricing for broader healthcare AI platforms.

The attraction is obvious: a single practice collecting $10 million annually can generate $300,000 to $500,000 in vendor revenue at a 3% to 5% fee. If AI significantly reduces the labor required to deliver that service, the potential margins are substantial.

Yet collections-based pricing has two weaknesses. Vendors earn more when customers collect more, even without improving billing performance, and providers may eventually question fees that rise much faster than processing costs.

The financial evidence also shows that alternative models already work at scale. Waystar generated $1.099 billion in 2025 revenue and a 42% adjusted EBITDA margin through a combination of subscriptions and volume-based services. Meanwhile, AI automation deployments are demonstrating that specific billing tasks can be handled with considerably less manual work.

For an AI company taking over the whole revenue cycle, percentage pricing remains a strong choice. For software that automates claims, eligibility, or payment processing, per-transaction fees and subscriptions are easier to defend. For denial recovery, charging against verified additional collections creates a clearer link to results.

We therefore favor a hybrid structure for the strongest long-term AI billing platforms: recurring fees for the core software, usage-based charges for routine processing, and performance fees where the vendor can prove it recovered additional revenue.

Percentage-of-collections pricing may generate the largest early contracts, but it won't automatically produce the most profitable or durable AI billing businesses. Those will be the companies that can automate expensive work, keep customers collecting efficiently, and charge fees that remain reasonable as both sides grow.

If you want more recent data on this point, please see our latest AI revenue cycle management market report.