Rob Ware still remembers entering healthcare revenue cycle management and hearing experienced professionals describe it almost as a form of secret knowledge — the rules were too numerous, payer behavior changed too often, and successful outcomes seemed to only smile upon those who had spent decades learning the exceptions.

“Everyone talked about RCM as if it was an art,” Ware reminisced. “Like you had to know all this magic — it was a black box.”

That realization was far from comfortable for someone who began his career in operations consulting, working in manufacturing and logistics. Ware had been trained to see systems, constraints, inputs, and measurable outputs.

Now, as senior vice president and general manager of revenue cycle management at ModMed, Ware believes artificial intelligence can finally address the variable-heavy complexity that has kept billing dependent on hard-to-scale human "magic." In a recent conversation on the MGMA Insights Podcast Network, hosted by Daniel Williams, senior editor at MGMA, Ware described an industry moving from retrospective denial analysis toward keeping claims from failing in the first place.

The "black box" meets a machine built for variables

Ware originally expected revenue cycle work to behave like a science: Follow the rules, submit the claim, and receive payment. Healthcare quickly corrected that assumption.

“What makes it complicated, from my perspective, is really about the number of variables and the fact that they’re constantly changing,” he said. Those variables include payers, benefit plans, coding combinations, medical-necessity requirements, authorization rules, and edits that can change without a practice immediately recognizing it's happening.

Humans can't reliably retain every combination and rule. “Even if you manually try to document and create the rules in your system, you’re always behind,” Ware said.

An algorithm can continuously evaluate more variables than an individual biller or coder and detect patterns across a larger claims population. However, Ware doesn't claim the industry has solved the problem. “Some of us are starting to show success,” he said, “but I would also argue no one’s fully cracked it yet.”

Those financial stakes are already visible. For example, MGMA’s January 2026 Stat poll found that 48% of practice leaders identified denials and appeals as their largest source of revenue leakage. 

When payers automate, practices can't answer monthly

Williams gave the competitive dynamic a science-fiction label: “fighting AI with AI.” But Ware’s vision is a bit less cinematic and more immediate need.

“The payers we know are using AI,” he said. “They are running first-pass denials with AI.” As payer edits become faster and more dynamic, practices relying on month-end or quarterly reports learn about a new denial pattern after rejected claims have already accumulated.

“We see this with our clients every day,” Ware said. “‘Hey, this was paying before. This is not paying now. What’s going on?’”

Historically, an RCM team would analyze the denials, identify the payer or code combination, create a rule, retrain staff, and monitor future claims. By the time the correction was in place, the payer could have introduced another edit. Ware sees AI as the practice’s best chance to detect the change quickly enough to alter the workflow.

Prediction is useful only while someone can act

The biller might need documentation from a physician who completed the visit days earlier, or coverage information from a patient who has already left the office. “When we talk about being preventative and predictive, it’s not just about the person who’s about to submit a claim,” Ware said. “They might need more information from the physician.”

His preferred model pushes the warning upstream: 

If a patient needs authorization, the front desk should know while the patient is scheduling or checking in. 

If documentation requires a specific clinical detail, the physician should receive that prompt during the encounter. 

If an insurance record is incomplete, staff should correct it before the patient leaves.

“What we want to do is actually bring that prediction all the way forward into the workflow to where you can do something about it,” Ware said.

This distinction should shape how practices evaluate AI products. A tool that predicts a denial but creates another queue for staff is an analytics layer. A tool that obtains missing information, routes an exception, or prevents submission until the issue is corrected changes the operation.

“People want more than prediction,” Ware said. “They want prevention.”

Buy against a leak, not an AI category

Ware’s purchasing advice begins with diagnosis. A practice should not buy “AI for RCM," but rather understand whether its measurable loss comes from charge capture, eligibility, prior authorization, coding, denials, or collections.

“I think that’s the one mistake I have seen — looking for something that just sort of solves a problem in general, as opposed to solving a specific problem,” he said.

Prior authorization offers a clear test case for evaluating technology: A practice should identify the exact breakdown, whether detecting the requirement, obtaining authorization, confirming coverage, or attaching the authorization to the claim, then select an integrated tool and measure its impact on authorization denials, clean claims, or staff workload. 

“If you’ve got something that has to exist outside and do a lot of data transfers, I think it’s just going to be tough for it to be as efficient,” Ware said. A collection of disconnected point solutions can replace one workflow problem with several interface problems.

Net collection rate is the end goal, not the early warning

Ware calls net collection rate the “ultimate barometer” of revenue cycle performance because it asks a direct question: Of the dollars contractually available to the practice, how many were collected?

“It’s out of what’s allowed, how much did I actually get?” he said.

Net collection rate accounts for contractual adjustments that make gross charges unreliable as a measure of performance. A practice operating at a 96% rate may no longer be able to treat the remaining 4% as uneconomical to pursue.

The drawback is timing. “It takes a while to get up to your 95%, 96%,” Ware said. “You’re probably a couple of months out.” Practices therefore need leading indicators, including eligibility-error rates, authorization failures, clean-claim rates, denial categories, and days in A/R, while preserving net collection rate as the final financial test.

Technology changes staffing, but not the need for judgment

Ware saw turnover intensify after COVID-19, especially in roles that require detailed knowledge of payer behavior.

Technology can reduce the amount of expertise required for every transaction by presenting relevant information to a less-experienced employee. “Some of this technology will enable you to potentially hire someone who doesn’t have 30 years of experience [but] can get up to speed faster,” Ware said.

When it comes to external RCM partners, practice leaders are often resistant because they fear losing the control available when the billing team sits down the hall. Technology can make outsourcing more palatable. Ware argues that real-time visibility, workflow tracking, and access to performance data can replace proximity with accountability.

Whether a practice automates internally or works with an RCM partner, the same prerequisite applies. “You can’t really fully harness the power of AI and some of these other technologies unless you’ve got clean information in a centralized place,” Ware said.

Ware’s relationship with technology has evolved with the tools themselves. For example, he once questioned the usefulness of the iPad. Today, when a new tool appears, he said, “I’m grabbing at them.” His daughter, meanwhile, reminds him that adoption is generational: “I use technology, but I still type, and my daughter does everything with voice.”

On Friday afternoons outside Boston, Ware steps away from the systems and variables for a hard mountain bike ride through public forests. The route isn't always mountainous, he admits, but it offers enough climbing to earn the fire pit and beer afterward. It's a fitting metaphor for someone trying to turn healthcare’s black box into a navigable trail: See the obstacle earlier, choose the right line, and keep moving. (Beer and fire pit optional.)

Resources

•  ModMed Revenue Cycle Management
• Connect with Rob on LinkedIn
• Latest revenue cycle resources at MGMA
• Beyond the hype: Practical AI applications that improve access, capacity, and revenue (MGMA article)
• How AI and Human Talent Can Work Together to Transform Medical Practices (MGMA podcast)
• Beyond the value-based care label: A framework for risk and payer contracting (MGMA article)
• MGMA Insights Podcast Network

Email us at dwilliams@mgma.com if you are an MGMA member who would like to appear on an episode. If you have a question about your practice that you would like us to answer, send an email to advisor@mgma.com. Don't forget to subscribe to our network wherever you get your podcasts!