Northeast Georgia Health System is accelerating AI adoption, especially in revenue cycle management, but leaders have learned to balance speed of innovation with ongoing ROI.
"You can chase every shiny object, but are you going to get the ROI?" asks Chris Paravate, chief information officer at Northeast Georgia Health System.
That shiny object, according to Paravate, is AI. The technology is everywhere in healthcare right now, but perhaps the best use case for applying AI to healthcare has been revenue cycle management. In this space, AI has been a transformative tool for automating the multitude of repetitive tasks throughout the revenue cycle, from registration and scheduling to denial management. And with such rapid adoption, AI is becoming a necessity rather than a nice-to-have.
That's why Northeast Georgia Health System (NGHS), a five-hospital, non-profit organization with more than 1,000 beds and 1,500 medical staff members, has made AI adoption a priority, particularly within the revenue cycle. The organization has already implemented over 120 AI applications, including its latest implementation of Epic's generative AI function to automate appeal backlogs.
And this is just the start of NGHS' AI adoption journey, Paravate emphasized.
"Our desire to go fast has really allowed us to accelerate some of these things," he told RevCycle Management. "It's not uncommon for us to say, 'Hey, we've got 30 or 40 projects in the queue, but oh my gosh, this is an opportunity. Let's move it to the front line.'"
AI in revenue cycle management is making a material impact. NGHS' Chief Revenue Officer Elyse Gates reported that the generative AI tool alone recovered over $3 million in additional reimbursements from a backlog of outpatient claims. She is already in talks with Epic as they develop similar technology to tackle inpatient appeals.
While recovering a couple of million dollars is certainly helpful for hospitals operating under razor-thin margins, any technology implementation needs to be part of a larger strategy to achieve true ROI.
"We want to be early adopters, but we also want to be strategic about not chasing every shiny object and turning everything on without really understanding where the benefit is and what ROI we think we are going to see from an improved productivity perspective," Gates said.
The trouble with KPIs
Revenue cycle officers like Gates rely on key performance indicators (KPIs) to assess their organizations' financial health. Revenue cycle KPIs track everything from days in accounts receivable and patient collection rates to denial rates and clean claim submissions. However, there aren't always established KPIs for painting a clear, complete picture of how an AI solution is generating ROI.
"It's not like we don't have KPIs that we track; I look at a hundred every single morning when I log onto my computer. But it's about tying those back to the AI solution," Gates explained. "You may look at your KPIs and see that net revenue is higher in an area, but you can't always tie that back to a solution. There could be several different reasons why you look better or worse."
Some examples Gates cited are changes in payer behaviors or employee workflows. These factors can influence the ROI generated -- or not -- after AI implementation. How much staff use the tool is also at play.
It's not like we don't have KPIs that we track; I look at a hundred every single morning when I log onto my computer. But it's about tying those back to the AI solution.
Elyse Gates, chief revenue officer, NGHS
Paravate uses a dashboard within the Epic system to track the AI applications turned on and how many queries staff make. For some applications, query volume spiked post-implementation, then dipped, raising the question: Is this solution not the right fit for NGHS' workflows or challenges? Or does the solution need to be tweaked?
For other AI implementations, like the outpatient appeals automation, query volume has remained high as staff continued to rely on the AI to gather the documentation needed for an appeal.
"We're seeing a usage count of about 3,000 per month out of that denials team, so it has a sustained value in aggregating that information," he explained. "What's harder for us to tell you is, 'Okay, so you ran that 3,000 times this last month, but how much time did it reduce in aggregating the information to file that appeal?'"
NGHS also needs to look at how AI has reduced full-time equivalents (FTEs), or more likely repurposed FTEs.
"Have we been able to get more records coded than we did previously?" Gates asked. "And that's not always the case depending on what you're turning on. So, there may be value there, but it may not be the value that you initially thought you were getting from it, either from a productivity or an ROI standpoint."
Identifying AI's value
Revenue cycle leaders must decide if AI solutions are delivering the value they need. Gates has decided to unplug AI solutions for revenue cycle management after the tool failed to deliver anticipated outcomes. In this case, the tool created more work for staff, rather than simplifying workflows, which disengaged the team. Gates stressed that AI solutions must make employees' lives easier and improve their productivity. If a tool fails to achieve this, it risks losing the team's support and enthusiasm.
It's important when talking about AI implementations, she stated, to have some established metrics to start and a good baseline of data to understand value over time. Revenue cycle leaders can then set expectations around improvement, whether they anticipate a ten or 20% yield. They must also have tools in place to track outcomes, not just after implementation, but on an ongoing basis.
The Epic dashboard is a good starting point for some of NGHS' AI solutions. However, Paravate is still seeking a standard set of benchmarks and KPIs for his organization's solutions. Gaining access to data on the performance of peer organizations would also benefit the health system's AI strategy, enabling it to optimize solutions for better results, he said.
"That's why we need that list to not just show usage," Gates added. "Everyone wants to know what we are getting out of a solution. So, we need to tie usage and adoption to outcomes, results, ROI and those types of metrics."
Balancing early adoption with sustained improvement
It's a balancing act, Paravate and Gates agreed. Healthcare organizations must balance the desire to innovate with the next big thing in revenue cycle AI with waiting on established KPIs to paint the full ROI picture.
"But if you wait until you have all the KPIs, then this thing could be gone," Paravate stated.
So, how are organizations successfully navigating this rapid pace of adoption while seeing results? Largely, with streamlined AI governance, according to Paravate.
"We see large health systems across the country, and their speed of adoption is largely tied to the complexity of their governance," he explained. "If they have complex AI governance, it can be the kiss of death; nothing is going to get implemented even though there's value there."
NGHS' board and CEO have made AI adoption throughout the organization a top priority, so speed has been top of mind for Paravate. This has led to a more streamlined version of AI governance that enables the new shiny object to prove itself.
If [large health systems] have complex AI governance, it can be the kiss of death.
Chris Paravate, chief information officer, NGHS
Paravate highlighted that at NGHS, an IT executive steering committee approves all AI initiatives. However, there is a tier beneath them that is split by business units aligned by customer groups, rather than by technology. This gives the revenue cycle a seat at the table with an IT leader like Gates to prioritize enhancements and changes specific to the revenue cycle tools within Epic, for example.
NGHS is taking AI governance a step further with a pilot of a third-party AI governance platform.
"It's similar to Epic's dashboard," Paravate explained. "It will show us a dashboard of all AI solutions, and it will look at what data those are accessing. Does it have appropriate access controls in place? What actions are being taken using that tool? What KPIs are being tracked?"
Streamlining AI governance is key as NGHS seeks to implement more AI applications. After all, a new form of AI for healthcare seems to emerge every day. Even within revenue cycle management, vendors and providers alike are now talking about applying agentic AI alongside generative AI, predictive analytics and robotic process automation.
"What happens when we go from 25 AI solutions we are using inside of Epic to 100 or even 200?" asked Paravate. "And that's just Epic, which is one of our over 400 applications that the IT team supports. We are going to have to have a digital mechanism to be able to manage and secure those AI solutions, and so we can monitor them effectively."
A future in which there are hundreds of AI agents working within the revenue cycle alongside staff is not far off. Healthcare providers and revenue cycle management vendors alike are heavily investing in AI to bolster their capabilities, if not create an autonomous revenue cycle using the technology.
However, AI adoption is not just about implementing the latest technology; it's about ensuring that each solution delivers measurable value and aligns with strategic goals. At NGHS, the balance between rapid innovation and thoughtful governance is paving the way for sustained improvement.
As healthcare organizations continue to navigate this transformative journey, the key will be to prioritize solutions that not only promise ROI but also enhance productivity and patient care in meaningful ways.
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.
