Most healthcare organizations are adopting AI in the revenue cycle: HFMA poll

An HFMA-FinThrive survey found that 63% of healthcare organizations use artificial intelligence (AI) and automation in the revenue cycle. This percentage will likely increase within the next 12 months as organizations look for ways to offset staffing shortages, increase cashflow and generate more revenue.

As healthcare organizations advance their RCM digital transformation, 63% have already integrated AI-powered automation solutions designed to streamline claims processing, enhance denial management and optimize overall revenue integrity. Of these, 15% have already seen a positive ROI, according to new research conducted by HFMA and FinThrive. The survey conducted between October and November 2024 includes responses from 101 unique healthcare organizations of various bed sizes and net patient revenues. Here are a few specific findings:

• 38% of organizations are doing groundwork or are considering pilot programs for selected AI technologies in the revenue cycle to reduce revenue leakage.
• 48% of organizations apply AI to documentation and coding. This early-stage adoption is proving to be the leading application, revolutionizing coding accuracy and operational efficiency. 
• 73% of organizations believe AI will have the biggest impact on prior authorizations to reduce administrative burden. 
• 67% of organizations recognize that AI and automation will drive the most significant impact on denials and underpayment management, accelerating recovery timelines. This area is also the leading focus for investment over the next 12 months. 
• 51% of organizations cite IT infrastructure limitation as the biggest obstacle to adopting AI and automation in the revenue cycle. Other top obstacles include lack of budget (44%), integration challenges with existing systems (43%), difficulty demonstrating ROI (42%) and vendor reliability concerns (42%).

“A couple of years ago, people rolled their eyes at AI just like they did at blockchain technologies,” said Jonathan Wiik, FHFMA, MBA, MHA, vice president of health insights at FinThrive. “Now, it’s all we’re talking about, and the conversations will only increase. If we redo this survey even as early as the next quarter, I think we’d see closer to 30% of organizations achieving a positive ROI.”

Choosing revenue cycle AI pilot programs wisely

While many organizations have already implemented AI technologies in the revenue cycle, others are embarking on pilot programs to determine efficacy, said Wiik. 

“AI is a ‘fail fast’ model,” he explained. “With that, organizations want to make sure the use cases are tested. It’s a large purchase, similar to buying a home. You’re shopping the neighborhood. AI is very similar to that. You’re shopping the tech, getting references and looking at multiple solutions. You want to make sure the application will benefit you.”

Set clear parameters. With pilot programs there is the need to identify clear parameters. Wiik points out that organizations must address these questions:

• What is the scope of the pilot program?
• What type of cases will the organization include?
• How long will the pilot program last?
• How will the organization define the pilot program’s success or failure? 

“With AI, you really should not try to boil the ocean,” said Wiik. “Instead, you should identify tasks in a pilot that are repeatable, predictable and measurable,” he added. “Focus on small, simple things first. Demonstrate effectiveness and then move forward. You’re looking for incremental process improvement.”

 Examine at least one of these metrics. Determining whether a pilot program should move beyond the initial phase to more widespread application requires organizations to examine one or more of the following metrics: 

• Discharge-not-final-billed rate: Can the organization submit claims faster?
• Cost to collect: Do claims require fewer human touches?
• Clean claim rate: Do claims have fewer errors? 
• Days in A/R or better yet, percent A/R over 90 days: Do payers pay claims more quickly?
• Net revenue: Do payers pay claims at a higher amount?

Organizations must also hold AI vendors accountable, said Wiik. 

“Ensure they can follow through in terms of what they say they can do for your organization,” he adds. 

 To promote this type of accountability, he says many organizations have begun to incorporate pilot programs into vendor contracts. 

“There’s a burn-in period of 90 days to ensure bots don’t fail and that data feeds script and populates fields correctly before everything is installed,” said Wiik.

Assessing the impact of AI on prior authorizations

While many organizations believe AI will have the biggest impact on prior authorizations, Wiik said pilot programs in this area can be tricky.

“It’s the one lever that payers can pull and move,” he added. “That’s a really dangerous area to apply AI in an overscoped way.”

In addition, there’s a lot of clinical variability with prior authorizations. For example, an organization obtains a prior authorization for an excisional biopsy. However, for clinical reasons, the physician performs an incisional biopsy instead. The claim may be denied even despite the presence of a prior authorization because the payer might require a reauthorization for the new procedure. Currently, AI can’t negotiate this scenario successfully, said Wiik.

“But the technology is getting smarter, and in the near future, AI may be able to determine why the doctor switched the procedure, explain that reason to the payer in a letter with relevant clinical and factual data, obtain a reauthorization and resubmit the claim, he said. “This can result in accelerating payments and reducing delays in care delivery.”

On the other hand, a more appropriate pilot program might be to determine whether AI can help with prior authorization requests that require supporting documentation. For example, an organization might leverage AI to identify when one payer requires additional documentation for patients undergoing a hip replacement, pull and attach that documentation and draft a letter to the payer requesting prior authorization. 

“This is a very easy, repeatable task,” said Wiik. “It’s easy to address this once you have the matrix, tables and rules in place.”

Another promising use case is using AI for denial management and appeals, said Wiik. AI can actively track authorization statuses and provide timely alerts for necessary actions. AI-powered tools can generate tailored draft appeal letters incorporating relevant clinical details, payer-specific guidelines and supporting documentation to reduce administrative burden.

Leveraging AI to identify underpayments

Denials and underpayments will be the largest area of investment in the next 12 months, primarily to reduce revenue leakage and to expedite denial and underpayment recovery timelines, the survey found. This includes investing in technologies that:

• Analyze the root causes of denials
• Automate flagging underpayment variances
• Draft clinical appeal letters
• Organize workflows for bulk appeals
• Prioritize appeal worklists

“Today’s organizations typically don’t have the capacity or time to look at every claim retrospectively, but AI can,” said Wiik. “It can flag claims that, based on historical data, appear to be underpaid and then dig into why. Was it a payer failing to recognize a contractual increase? Downcoding? Prior authorization issue? Line-item variances or denials?”

It’s always important for organizations to identify underpayments; however, Wiik suspects the technology’s real value proposition for organizations will be the ability to identify patterns of underpayments. 

“Payers are accustomed to strong-
arming providers into settling underpayments, but that will change as providers become available to identify widespread patterns of underpayments and become able to make a broader case for being paid what they deserve,” he added.

Achieving a positive ROI of AI in the revenue cycle

For now, many organizations continue to see a positive ROI with AI technologies in the revenue cycle, said Wiik.

For example, AI can streamline administrative complexity in the revenue cycle, said Wiik. 

“The prior authorization process is manual, time-consuming and highly variable, requiring staff to navigate payer-specific requirements, gather supporting documentation and submit requests — all while managing frequent delays and denials,” he said. 

“A single missing document can result in claim rejections, leading to costly rework and delayed patient care. AI improves approval rates and minimizes denials by automating document retrieval, identifying payer requirements and generating complete authorization requests. This reduces administrative burdens, allowing staff to focus on higher-value tasks while preventing revenue leakage. Faster approvals also enhance patient access to care, leading to better outcomes and greater financial efficiency.”

The appeal process for overturning denials and underpayments that create significant financial burdens for healthcare organizations could be another prime area for AI-driven optimization, said Wiik. 

“Manual processes for identifying, appealing and preventing denials are resource-intensive and prone to errors, leading to revenue leakage and operational inefficiencies,” Wiik said.

“AI enhances ROI by automating denial management and analyzing patterns to predict and prevent future denials and underpayments,” he added. “AI-driven payment reconciliation tools can detect underpayments, flag discrepancies and initiate corrective actions, ensuring organizations receive full reimbursement. By reducing manual rework, accelerating appeals and improving cash flow, AI helps healthcare providers recover lost revenue, lower administrative costs and enhance overall financial performance.” 

AI can also increase capacity, throughput and yield while also reducing labor costs. 

“I know lots of organizations that are auto-coding diagnostic tests like lab draws, X-rays, CTs and MRIs,” Wiik said. “Claims are going out the door within hours of the studies instead of within several days if a coder had to do it. This is good because it’s significantly faster, coders aren’t cheap and there’s not a lot of them.”

In addition, ambient documentation technology allows organizations to identify missing diagnoses and improve net revenue while simultaneously improving physician satisfaction. 

When thinking about ROI, it’s important to differentiate between hard savings (i.e., truly earned savings or cash) versus soft savings (i.e., savings that won’t likely hit the balance sheet), said Wiik. Equally important is the ability to obtain baseline data to show improvement, he added.

Overcoming persistent barriers to adopting AI in the revenue cycle

As organizations strive to overcome barriers to AI in the revenue cycle, Wiik says a robust IT infrastructure allowing easy data ingestion will be critically important. 

“Having eight instances of the same [EHR] won’t allow data to flow seamlessly, and adding AI on this might not be effective,” he added.

Securing IT talent is also critical in terms of ensuring organizations have the right in-house knowledge and skills to support AI technologies.

AI governance will be equally important. He says organizations must be prepared to answer these and other questions as industry standards and regulations evolve: 

• How will the organization promote data privacy and security?
• Who owns the data? For example, with an eligibility transaction or cost estimate, is it the payer? Provider? Vendor?
• What is the contingency plan if the AI technology fails suddenly?
• How will the organization ensure accuracy? What if the AI omits a diagnosis or flags a claim for review inappropriately?

“I think we’re going to see a lot of accountability for the hospital both clinically and financially and that can be a barrier,” said Wiik. “How much of this do you want to open and run without guardrails in place?”

While some organizations believe that AI is cost prohibitive, Wiik said this argument is becoming less of a barrier.

“Margins were predominantly negative a year ago. They’re predominantly positive now,” he adds.

Conclusion

There are many areas where AI technologies have shown significant progress, and while adoption barriers persist, the industry continues to move toward intelligent automation in the revenue cycle. 

“It’s an exciting time,” said Wiik. “People aren’t as afraid of AI as they used to be. They want it now because they know they need it now. We’re seeing a lot of trust in and adoption of AI now than we’ve seen in the last decade. People are using it and seeing the benefits of it. Now the question is how to organically grow it in ways that are safe, applicable and fair? And more importantly, how do we use it in ways that ultimately benefit patients?”

About FinThrive

FinThrive is advancing the healthcare economy. Our 1,600-plus colleagues rethink revenue management to pave the way for a healthcare system that ensures every transaction and patient experience is addressed holistically. We’re making breakthroughs in technology—developing award-winning revenue management solutions that adapt with healthcare professionals, freeing providers and payers from complexity and inefficiency, so they can focus on doing their best work. Our end-to-end revenue management platform delivers a smarter, smoother revenue experience that increases revenue, reduces costs, expands cash collections, and ensures regulatory compliance. We’ve delivered over $10 billion in net revenue and cash to more than 3,245 customers worldwide. When healthcare finance becomes effortless, the boundaries of what’s possible in healthcare expand. For more information on our new vision for healthcare revenue management, visit finthrive.com.

_{This published piece is provided solely for informational purposes. HFMA does not endorse the published material or warrant or guarantee its accuracy. The statements and opinions by participants are those of the participants and not those of HFMA. References to commercial manufacturers, vendors, products, or services that may appear do not constitute endorsements by HFMA.}