While revenue cycle leaders at many health systems say artificial intelligence is providing real, measurable value, the largest wins in its journey have come from targeted automation instead of sweeping transformation.
From coding assistance to denial management workflows, AI has proven most successful when embedded into existing processes rather than stacked on top of them. However, its limitations are equally relevant, with prior authorization bottlenecks, patient-facing registration tools and analytics platforms moving slower than anticipated. The growing consensus: As AI’s ceiling in revenue cycle is still being established, the floor is rising, and teams that see the best results are investing as much in workforce preparedness as they are in the tools themselves.
Becker’s connected with Blake Evans, System Vice President of Revenue Cycle, Chicago-based Rush University System for Health; Mark Cannon, Vice President of Revenue Cycle Management, King of Prussia, Pa.-based Universal Health Services; Carrie Donovan, SVP, Chief Revenue Cycle Officer, Charlotte, N.C.-based Advocate Health; and Lissa Mann, VP of Revenue Cycle Apps, Advocate Health, to discuss how AI is reshaping revenue cycle operations across their systems.
Editor’s note: Responses have been lightly edited for clarity and length.
Question: Where have you seen AI make the biggest measurable impact in your RCM operations? What’s the specific metric that convinced you it was working?
Blake Evans: We were experiencing a significant rise in request for information and additional documentation denials from one of our largest payers. A key challenge was that the payer does not allow submission of supporting records at the time of billing, creating delays and additional manual work once denials were received.
In response, we partnered closely with our internal AI team to design and implement an automated workflow that triggers an immediate response to RFI denials upon receipt of the electronic remittance. This approach has significantly reduced turnaround time, improved operational efficiency by reallocating staff to higher-value work, and provided greater visibility into denial overturn rates once documentation is submitted.
Based on this success, we are actively expanding the automation across additional payers and scaling it to both hospital billing and professional (physician billing) workflows.
Mark Cannon: AI has made a significant impact in our ability to triage and appeal insurance denials, especially ones requiring complex decision-making and access to a variety of information (e.g., managed care agreements, payer policies, medical records, etc.). Over the past several years, many payers have increased both “soft” denials (e.g., medical record requests) and “hard” denials (e.g., no authorization), and with tight timely filing windows and rising collections costs, using AI is a natural progression for hospitals and other healthcare providers. The main metric for this is staff efficiency, specifically the volume of claims or denials requiring follow-up that are processed daily.
Carrie Donovan: I think we were out of the gate first in our mid-revenue cycle with respect to coding. So we’ve taken advantage of both Epic solutions, as well as solutions offered in some of our vendor software. It’s measurable, because we can clearly count the number of encounters or sessions that are addressed, and so we can measure that against our own manual efforts to articulate that into real cost savings.
Lissa Mann: We’ve seen the biggest measurable impact where AI is directly embedded in the workflow. In the coding example that Carrie mentioned, we have scaled that to all of our workflows for coders, and we’re seeing improvement in both coder productivity and a reduction in coder related denials using those tools. That really has been a tremendous value add for us across the enterprise.
Q: Which AI use case were you most optimistic about that hasn’t delivered? What did you learn from it?
BE: Securing prior authorizations before the date of service remains one of the most administratively burdensome aspects of the revenue cycle. To address this, we’ve made meaningful progress by centralizing our authorization teams, standardizing workflows and optimizing the use of payer portals.
We were encouraged by the introduction of Epic’s payer platform, which offers the potential to streamline connectivity between providers and payers. Our goal was to leverage this functionality to automate authorization requests and receive approvals with minimal manual intervention. While we have seen early success, adoption has been limited to a small subset of participating payers, representing only about 7% of our total authorization volume.
As a result, we are evolving our strategy. We will continue to utilize Epic’s internal electronic prior authorization workflows while also partnering with a vendor to expand connectivity across a broader set of payers. Through this combined approach, we expect to significantly increase our electronic authorization rate and reduce manual effort over the coming months.
MC: Integrating AI into patient interactions at registration to improve insurance capture and financial clearance has proven difficult, especially for emergency room visits. AI requires near real-time access to several systems, including the electronic medical record, integration with clearinghouses or payer portals, and workflows that are convenient for both patients and staff. While we believe we’ll continue to evolve the use of AI in this space over time, we learned the importance of both centralized data management and end-user testing throughout the development process. We’re also still limited by payer access points such as portals requiring manual and unique logins and costly connections with other electronic data companies.
CD: We’re looking to advance AI, and truly across the revenue cycle, there is a lot of automation embedded, but there hasn’t been large-scale case use that really replaces the work of a whole lot of people. It’s not really that it failed, but maybe it didn’t achieve the lift that we’re really looking for to complement our efficiency. In more recent times, we’re starting to see more impactful solutions. But, if we look back over the last couple of years, it’s really been just small bites. Even networking with my revenue cycle peers, nobody’s got the one big thing that has really made a meaningful difference.
LM: It’s oftentimes when AI doesn’t work, sometimes that’s actually a human problem. Something that we’ve been more cautious on has been AI-driven insights and analytics. It’s something that we have rolled out and we are using, but the value comes from how we’re using that. The insights alone aren’t creating value, it’s that execution and getting it embedded into workflows, making sure we have clear ownership of how that informs our processes. It’s not that the AI doesn’t work, but, are your teams ready? Do they have the literacy for it? Can they really use those tools in that next level? I think that’s where we’re still seeing development in that space and working on upskilling our teams to be able to use that at a higher level of analytics.
Q: How has AI changed the role of your top RCM performers? Are they doing less, or are they doing something fundamentally different?
BE: The work of our top performers hasn’t fundamentally changed, but our mindset has. In our revenue cycle leadership discussions, the starting point is now: How can automation help solve this?
We’ve seen a clear shift in how our teams approach problem-solving. Instead of maintaining the status quo or protecting legacy workflows, there’s a growing sense of curiosity and openness to change. We are rethinking previous problem areas and questioning constraints that were once accepted as fixed. The focus has moved toward continuously identifying opportunities to simplify, streamline, and make processes more efficient through innovation and automation.
MC: AI is eliminating non-value-added administrative tasks. No longer do staff need to spend time opening a variety of systems, reviewing hundreds of notes, uploading documents to portals (or worse, preparing physical mail) and distributing worklists. Registrars now spend more time talking with patients about financial assistance options; appeals nurses work at the top of their licenses to review clinical denials and focus on performance improvement and improved care coordination with payers; and CDI specialists know they’re tackling the highest-value documentation and coding opportunities with our providers. The switch to AI has maintained a “human in the loop” for every process, so I think of it as lift-and-shift to higher value work versus doing less.
CD: We’re really invested in upskilling and preparing our teammates for the future with AI more deeply embedded into our workflows. In an organization of our size, we never get to the bottom of the work queue, so these accelerants really help us to be more productive. We really haven’t gotten to a space where we’re nearly out of work. It’s really important to engage our frontline teammates, as well as our leaders, for their ideas of ways to deploy the AI, but also where to focus their energies when the solutions make us more efficient.
LM: Something that we hear peers say in clinical spaces often is, we want to make sure everybody’s working with the right licensure, and we’ve taken that to heart in the revenue cycle space too. We’re making sure that as we’re giving our teammates these tools that make them better at the work that they’re doing, not only to address that volume, but talking about our coding example, we’ve had folks say, “Oh my gosh, it’s like having a second set of eyes.” Now they’re working by exception. They’re using good judgment. They’re able to validate and improve those AI outputs and really inform the solutions and then optimize the end-to-end processes. It is challenging them to work at the top of their licensure, and improving our end-to-end outputs too, by being able to see the big picture within the revenue cycle. That has been a tremendous improvement, and our teammates have really taken to the adoption of our AI tools because of that. We want their voice in the workflows. We want to know what we could be doing differently, and now they have the capacity to share that information with us. They’re not doing things like data entry or routine work. They’re working on those highest priority items, and they’re able to say, “Hey, Lissa, we have a different problem here.” They can do some of that root cause analysis, they can surface those insights to us. It’s changing the way that we work, but we’re finding it’s giving us higher value opportunities to connect and collaborate.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
