Two to three years ago, investors were telling startups that clinical coding would be fully automated by LLMs within a year — which didn’t quite turn out to be true.
Healthcare’s revenue cycle has turned out to be stubbornly resistant to easy technological fixes, but AI can still make real progress, according to Lee Kupferman, Co-CEO of R1’s innovation lab.
AI makes the most impact when it’s used to handle straightforward, high-volume work, freeing humans for the cases that actually need their expertise, Kupferman noted during an interview last week at HFMA’s annual conference in National Harbor, Maryland.
For instance, a simple inpatient encounter where a patient came in for a known procedure with no complications is the kind of case where 50 coders would all reach the exact same answer — that’s where AI should just run, he said.
However, most AI models still struggle with complex encounters that involve longer documentation and more varied payer rules. Kupferman believes the goal right now is figuring out how to route the right work to AI and reserve human efforts for the gray areas.
“You can get value out of [AI] tools in all of the revenue cycle, provided you have the right guardrails and you’re honest about where it works well and where it’s still got a way to go,” he remarked.
Part of what makes the revenue cycle resistant to AI is that the healthcare payment system itself is so deeply fragmented, Kupferman added.
There are hundreds of vendors within the healthcare revenue cycle space, but most of them sell narrow point solutions that don’t communicate with each other, he pointed out.
Health system coding teams often operate in near-total isolation from the prior authorization team, which means a denial that could have been caught upfront triggers weeks of rework downstream instead, Kupferman explained.
He views this fragmentation as one of the biggest obstacles to AI actually delivering on its promises. Kupferman said these tools need to be connected in order for efficiency gains to actually be realized.
The good news, though, is that the environment might be changing. Kupferman noted that while payers and providers have both historically dismissed the idea of a more collaborative, AI-driven revenue cycle, they’re starting to show more willingness to work together on improving the payments process.
“Everybody is in violent agreement about what the problem is — they’re just trying to figure out the best way to solve it,” he declared.
Photo: uchar, Getty Images
