This article is sponsored by Joyful Health. This article is based on a Behavioral Health News discussion with Becky Carlson, Head of RCM at Joyful Health, Emma Sugerman, Co-founder and COO at Mavida Health and Arnav Simha, Lead PM at Nirvana. The discussion took place on March 5th, 2026 during the BHB VALUE Conference. The article below has been edited for length and clarity.
Behavioral Health Business: Automation in revenue cycle management has existed for years. What fundamentally is different about this current wave of AI, and where is it actually changing outcomes rather than just improving efficiency? Becky, why don’t you kick that one off?
Becky Carlson: So I think automation in RCM, as you mentioned, has been around for a long time. We’ve seen auto-adjudication systems on the payer side forever. Where things are different now with AI in RCM is the reasoning about the data. We’re seeing opportunities to improve how we predict whether claims are going to be paid. We’re also seeing the ability to better prioritize claims based on that reasoning, so we can start to understand which claims will deny before we even take action on them. So it’s really about adjusting the framework—not just doing the activities, but changing the order in which we do them, how we prioritize them, and the actions we’re able to take.
BHB: Emma, what are you seeing?
Emma Sugerman: Yeah, I think we have smarter tools now, right? But they’re also easier to use, and that’s really important because there’s still a human component here. We have teams doing this work and contributing to it. We’ve had tools for some time that have helped, but it’s never quite been where we are now. It’s never quite been enough in terms of having the right information at your fingertips in a way that’s easily digestible so you can make the next right decision. That’s the shift we’re seeing now.
Before we partnered with Nirvana at Mavida, we had the information. It wasn’t that we didn’t have a tool, but the information wasn’t actionable or easy to understand. So with the same team—what I like to call the same operational horsepower, manpower, or woman power, as it tends to be for us—we can do a lot more. And that translates to margin improvement and operational efficiency.
BHB: Yeah, that makes a lot of sense. Arnav, what are you seeing?
Arnav Simha: From the perspective of building products, there are really two angles of technological advancement that are important to note. First, companies like Nirvana—companies focused on AI—can continue to be agile and adopt the latest technology. But we’re only as good as the systems providers actually use.
As technology has advanced, EHRs and practice management systems have improved as well. Five years ago, those systems were extremely difficult to work with. Even if there was a phenomenal AI product on the market, it was hard to meet providers where they were because the connection to those systems was essentially broken.
Now we’re seeing what I’d describe as a rising tide lifting all boats. The technology within EHRs is improving, which allows companies like Nirvana to integrate AI agents in useful ways—not only to gather information more effectively, but also to deliver that information back into the right place at the right time.
BHB: Where are organizations overestimating AI’s impact in revenue cycle, and what has to exist before it truly delivers ROI? Emma, do you want to kick that off?
Sugerman: Sure. For me, it comes back to thinking about how the tool will actually be used and the context in which it’s used. There’s also change management to consider. There’s often resistance to new technology, and that means we need to train people and help them understand it.
Being mindful of those layers—how this technology impacts an organization and what the onboarding process looks like—is really important. If that resistance persists, we won’t see the impact or improvements we expect.
Carlson: I think some people view AI as a magic wand that’s going to save RCM. The reality is that if your RCM processes are broken and you introduce AI, you’re just going to get chaos faster. So it’s important to make sure you’re fixing the core issues first. AI should be used as a workforce enabler—a force multiplier that increases the amount of work a team member can accomplish. But it still requires cleaning up processes and making sure your RCM operations are solid before introducing it.
Simha: I take a similar perspective. You’ll hear a lot of people say AI is a magic wand for everything. It absolutely isn’t, and it’s not the right way to think about integrating AI.
The right approach is crawl, then walk, then run. In the crawl phase, you look at the volume of information you’re already dealing with—claims, checks, portal interactions—and identify the biggest problem areas. In healthcare, the information is often there. The issue is that it takes too long to access or requires too much manual work.
So you start with the processes that take the longest time, require phone calls, or produce unreliable information. That might only be five to ten percent of your volume, but they’re the biggest strain on your workforce. If AI can reliably solve those problems, then you can start expanding from there.
BHB: How should leaders rethink prioritization in an AI-enabled revenue cycle, and what changes when AI informs which claims get worked on first?
Carlson: Traditionally in RCM, we’ve prioritized claims based on aging buckets. You go to conferences and everyone talks about zero to 30 days, 31 to 60, 61 to 90, and so on. The problem is that the likelihood of collecting in each bucket is different, and the level of effort required is different.
RCM produces enormous amounts of data, but very few organizations use that data as a feedback mechanism for future decisions. By leveraging AI, we can identify patterns within claims—claims that have a high likelihood of being resolved successfully, and claims that unfortunately are unlikely to be paid.
That allows organizations to make faster decisions about where to focus their time. When you’re dealing with so many claims and so much data, that kind of prioritization becomes incredibly valuable.
BHB: If an organization is still prioritizing claims purely by aging, what risks are introduced?
Carlson: There are many different ways to approach prioritization when you have strong feedback mechanisms. For example, if you’re working claims strictly by aging and there’s a payer enrollment issue affecting your organization, your team might spend a lot of time navigating the noise instead of fixing the root problem.
Once that enrollment issue is corrected, many of those claims could be reworked automatically. But when you’re focused exclusively on aging, you might end up working on easier claims first while more complex issues continue aging out or falling through the cracks. Looking at things like payer mix and prioritizing the payers that represent the largest portion of your revenue can be a much more strategic approach.
BHB: When you implemented eligibility-focused AI at Mavida, what operational frictions were you trying to eliminate, and how did it change how you think about scaling revenue operations?
Sugerman: We had a tool before we worked with Nirvana—it just wasn’t getting us where we needed to be. One major issue was that the information coming back from eligibility checks wasn’t digestible. It wasn’t actionable for our team or for me.
So we had information that we were trying to translate to patients, but it wasn’t clear enough for anyone to confidently act on. Once we implemented smarter tools, the difference was immediate. The information became easier to understand, easier to communicate, and easier to act on. That allows each operator on the team to accomplish a lot more.
BHB: What didn’t improve immediately, or required more operational change than you expected?
Sugerman: For me it comes back to the idea that the inputs determine the outputs—what people often call “slop in, slop out.” If the underlying information isn’t complete or standardized, the tools can only do so much.
For example, we could verify that a patient’s plan was valid and active, but answering bigger questions—like whether care is in-network and what the patient’s financial responsibility will be—is still challenging. The worst experience for a patient is receiving care and then getting a surprise bill later. So there’s still work to do across the industry to improve the accuracy and clarity of that information.
BHB: Arnav, from a product perspective, what has to be true inside a revenue organization for AI to consistently perform?
Simha: The essence of “slop in, slop out” is that the data feeding the AI needs to be coherent. AI won’t be very effective if it has to interpret scattered information across notes, fields, and documents.
Organizations need to normalize their data so AI systems can clearly understand the task they’re performing and the output they need to provide. Ideally, AI vendors should help standardize that information—from the data in your systems to the outputs written back into them. The cleaner and more structured the data is, the more effective the AI will be.
BHB: When AI underperforms in the revenue cycle, is it usually a product issue or an environmental issue?
Simha: It’s usually both. AI tools should be treated the same way you’d train a junior employee. You test them, they make mistakes, and you provide feedback.
Some of that feedback might involve improving how data is structured within your EHR or practice management system. Other times, the AI model itself needs refinement. By identifying failure cases and categorizing them, organizations can improve both the environment and the product until the system becomes a highly effective contributor.
BHB: If someone here is evaluating an AI solution tomorrow, what’s the first question they should ask?
Carlson: The most important question is: what job will this AI solution actually do? And what metric will measure its success?
Are you trying to improve eligibility checks, denial management, or coding? Once you define that, you need a clear metric—like reducing eligibility-related denials, improving denial overturn rates, or ensuring documentation passes compliance review. Being intentional about the specific job and the success metric is critical.
Sugerman: For me the question is: how will this supercharge my company and my team? I’m thinking about margin improvement, operational efficiency, and how we can make the same resources go further. The tool doesn’t stand on its own—it has to amplify the team’s capabilities.
Simha: I think there are two questions. First: how does this help my organization scale? You need to understand where your current metrics are and where they need to be to achieve your goals.
Second: what job is the tool actually solving? Every AI tool is essentially performing a specific task—making a phone call, reading a portal, interpreting notes. You need to determine whether that task represents a meaningful pain point for your team.
BHB: Looking ahead, what are you most excited about for the future of this technology?
Carlson: After spending a decade in RCM watching teams spend huge amounts of time just trying to identify problems, I’m excited about the ability to analyze large volumes of data and identify patterns earlier. Instead of constantly reacting to issues, organizations can anticipate them and address them proactively.
Simha: The future we’re excited about at Nirvana is one where healthcare transactions are as simple as buying a bottle of water at a deli. You walk in knowing the price, pay with a trusted method, and receive the service.
Today, bad or incomplete information makes that difficult. But AI can identify trends across claims data and uncover patterns—like identifying when claims are consistently rerouted to a different payer. That allows organizations to make faster, more confident decisions and dramatically reduce the time it takes to resolve transactions.
Sugerman: For me, it comes back to access to care. At Mavida, we provide deeply specialized care. Patients want to know their care is in-network, affordable, and predictable.
I’m also excited about leveling the playing field for providers. These tools shouldn’t only be accessible to very large organizations. If we can democratize access to these technologies, more providers can accept insurance, which ultimately expands access to care for more people.
Joyful Health helps healthcare practices identify and reclaim lost revenue. Their revenue intelligence software shows exactly what you’re owed and where you might be leaving money on the table, and their expert team goes to bat to recover any missing insurance revenue for you. To learn more, visit: https://www.joyfulhealth.com/.