The next big thing: Agentic AI for revenue cycle management

What is agentic AI for revenue cycle management?

Last year, the buzzword was generative AI; the year before, it was machine learning. Now, agentic AI is taking center stage. But how does it differ from other forms of AI, and what makes agentic AI poised to disrupt revenue cycle management in ways other technologies haven't?

"When we talk about agentic AI overall, it is similar to other workflow technology in the past, but it is autonomous," explained John Landy, chief technology officer at FinThrive. "It allows for decision-making, and it's integrated. So, by nature of it being in the AI realm, which is broad in general, it is integrated with data systems."

In comparison, other AI or automation capabilities, like robotic process automation (RPA), are deterministic. In other words, these capabilities rely on rules to predict an outcome, and that output will always be the same given the input.

Agentic AI is probabilistic, reflecting its ability to assess patterns in data and adapt to new information and conditions. Generative AI is another example of probabilistic AI. However, agentic AI is an evolution in the sense that it can operate autonomously and make decisions.

These capabilities make agentic more capable of handling complex tasks versus generative AI, which experienced significant growth in revenue cycle management over the last two years.

For example, generative AI can write an appeal for a denied claim after some prompting, but agentic AI can identify a denied claim on its own, analyze why it was denied, gather appropriate documentation, write an appeal, and even submit that appeal to the payer. Agentic AI can also do this with minimal human intervention, whereas generative AI requires prompting and does not take action beyond what it is prompted to do.

Why now?

A reasoning, digital assistant is exactly what revenue cycles need in this environment, according to Dan Parsons, chief experience officer and co-founder of Thoughtful AI, a provider of AI-driven revenue cycle management solutions.

"The thesis underpinning the application of agentic AI is a couple of macro trends that we see directly impacting revenue cycle -- that perfect storm. It makes the 'why now' very evident," Parsons explained. "One is the staffing shortage. As we think about the different staffing components required to support the complexities of the revenue cycle, every provider we talk to is either understaffed due to increasing costs or unsure how to maintain the level of staffing required to support the increased volumes."

The situation is further complicated by the complexities within the revenue cycle, which generally includes a front, middle and back to complete the medical billing process. These sections also tend to be siloed despite their interconnectedness.

Staff must also contend with dozens of payer contracts, each with its own set of rules and requirements for reimbursement.

"This leaves finance and revenue cycle leaders really stuck in a spot where there's no great answer," Parsons added. "The traditional methods of scaling your in-house team just aren't working. The talent pool isn't there. Younger generations don't want to do this work. And the rising labor costs make it increasingly harder for traditional outsourcing methods to be cost-effective, especially as we look at the margin profile of a lot of these organizations being 1 to 3%. There's just not a lot of room to breathe."

How agentic AI will solve revenue cycle challenges

Agentic AI presents an opportunity to tackle these revenue cycle challenges in ways that other AI solutions couldn't.

"You start to apply these very modern capabilities that really only became generally available in the last 24 months or so, even at some key points across a very comprehensive revenue cycle workflow, and you start to get a ton of leverage," Parsons said.

Healthcare organizations haven't necessarily gotten the same leverage from other types of automation or AI.

"As we think about GenAI or even RPA, there were a lot of reasons why that fell short and the promise wasn't ultimately delivered," Parsons stated. "A lot of it just came down to the technologies being too brittle, and it being too costly to maintain."

However, healthcare organizations have come a long way since the days of implementing RPA and even generative AI. Healthcare organizations are becoming increasingly sophisticated with their data, infrastructure and technology use. A rich history of using AI in revenue cycle is paving the way for easier adoption of newer, more mature capabilities like agentic AI.

"We needed to let the technology and just the pure computing costs and speed catch up before we could introduce it in a production scenario in a safe way," Parsons added.

Now, agentic AI is poised to uniquely address revenue cycle challenges by combining autonomous decision-making with self-learning capabilities, while orchestrating across multiple complex healthcare processes in a way that traditional automation can't achieve.

"The orchestration layer is key," said Judson Ivy, founder and CEO of Ensemble Health Partners. "It doesn't matter whether the agent comes from Microsoft or if it's native to the EMR, such as Epic. It doesn't matter whether the agent is an Ensemble agent that we've created. What is critical is that they all speak to each other and learn from each other so we can solve the friction in the revenue cycle versus reinforcing all those verticals within it."

Supporting this orchestration of AI agents across the revenue cycle can lead to a more efficient revenue cycle, not just improvements in one or several KPIs, according to Ivy.

"If you think about the revenue cycle prior to agentic, the same problem exists. You have very smart coders who would run into issues because the person who registered the patient or who entered the charges made a mistake. Just because you identify a particular function to apply AI to doesn't mean those upstream or downstream problems don't exist," Ivy explained.

Orchestration addresses this issue through AI agents that can communicate with each other to identify workflow breakdowns. This also supports the use of AI at scale, according to Ivy.

Agentic AI may be the latest buzzword for healthcare finance and revenue cycle leaders, but it could also be the next disruptor in a space ripe for disruption. Adding AI agents to the revenue cycle toolbox can not only further automate an otherwise largely manual and repetitive process but also push toward a more intelligent revenue cycle.

Jacqueline LaPointe is a graduate of Brandeis University and King's College London. She has been writing about healthcare finance and revenue cycle management since 2016.