I hate how hard it is for doctors to get paid today. I know I’m not alone. I doubt there are any physicians or practice managers in the U.S. who feel like this current process of charting, coding, denials, quality measurements, and clawbacks is working well.
It’s still worth saying out loud though: I hate it. You probably hate it, too. The whole “revenue cycle management” world that we’ve come to see as normal is not normal. It’s bad.
The complexity involved in insurance billing repeatedly astounds me. Whether fee-for-service or value-based, the additional task load required of clinicians in order to get fairly paid for the work they do is unlike any other reimbursement practice in any other industry. I’m saying this as someone who lived it as a front-line biller in a medical practice and who studied it as a health economist: it’s madness.
In primary care, I’d posit we have it even worse than other specialties. Our team has become deeply familiar with Advanced Primary Care Management (APCM), a CMS program launched in 2025 for Medicare patients in a per member per month model, and no wonder adoption is abysmal. It’s labor-intensive, complex, and EHRs aren’t built to automate away the toil. But things might be changing.
AI is here, and it is going way beyond scribes. AI is showing it might be terrific at taking certain dauntingly complex processes and whisking them into the background, if we can get it right. If poorly designed, AI could generate low-quality bills with errors needing correction before or after denials and even more “pajama time” for practices.
If well designed, it could be a game changer. At Elation, we approached AI billing with two key priorities: we needed it to smartly reference existing clinical and claims data already in the EHR, and to excel with the most common, highest-volume work, not to attempt to be a one-sized solution for every single bill. We created a system that uses what’s already in the EHR, pre-visit details, visit notes, the problem list, orders, medications, practice and insurer billing history, to suggest more complete diagnosis, procedure, and drug codes before a bill is created. And then we sorted the easiest bills into a touchless fast lane for submission, leaving complex exceptions out and clearly flagged for the practice’s manual review.
In our pilots, we’ve seen more than 70 percent of eligible claims created touchlessly from visit note sign-off. Of those, more than 95 percent were reviewed, confirmed, and submitted for reimbursement without any additional manual intervention. That’s nearly 7 out of 10 bills we’ve made easy for AI to manage on its own, and the remaining 3 supported by humans. Less fuss with a goal of more quality, and greater reimbursement. We’ll be watching closely and working furiously over the next few months to keep raising these numbers. Still, even with removing a huge administrative burden for most bills from physician practices, this is a small step toward ending the overall billing madness.
Where things go next could be way more interesting. Could future AI be trusted to simply tell payers what they owe you, without the bills and codes? What if AI could smartly identify the complexity of your patient panel and the quality of your clinical expertise to negotiate with payers over your rates and bonuses? Maybe AI gets so good at identifying actual fraud and compliance patterns that it changes how we think about patient documentation and who it is for? Could our national army of human coders be transformed into new roles such as medical librarians, patient care coordinators, or community health workers?
My hope for AI is that it eventually leads us to a health care payment future that we don’t all hate. In the meantime, we all need to keep chipping away at making the existing system less painful for clinicians and their teams.
Kyna Fong is a health care executive.
