Agentic AI Is Quietly Eating the Revenue Cycle
The flashiest AI demos are about diagnosis. The real money is in billing and coding. Here are the four no-regret bets, and the second-order effects nobody demos.
Photo by Tara Winstead on Pexels
Agentic AI is landing first and hardest in the revenue cycle, the billing, coding, and prior-authorization back office, because it is the one corner of healthcare that is structured, repetitive, and tied to clear dollars. Roughly 63 percent of organizations have already integrated AI into revenue cycle workflows and 80 percent are exploring generative AI for it, a 38-point jump in under two years, with ambient documentation, clinical documentation improvement, coding, and electronic prior authorization as the four no-regret bets. The catch is second-order: a payer-provider AI arms race that mostly shifts cost rather than improving care, a widening gap between large and community providers, and EHR vendors poised to absorb the point solutions.
Everybody wants to talk about AI that diagnoses cancer. Meanwhile the real revolution is happening in the least glamorous room in the building: the billing office. It is not sexy, nobody makes a heartwarming ad about prior authorization, and that is exactly why it is where agentic AI is winning right now. If you want to see where this technology is actually earning its keep in healthcare, follow the money, not the magic.
The least sexy corner of healthcare is where AI actually works
Think about what agentic AI is actually good at: structured, repetitive tasks with a checkable right answer and low harm if it slips. Now look at the revenue cycle. It runs on codes, rules, and forms. Every claim comes back paid or denied, so you always know the score. A mistake means rework, not a hurt patient. And the ROI shows up in dollars this quarter, not in a hard-to-price outcome five years out. Diagnosis is the opposite on every axis. That is not a knock on clinical AI, it is just why the billing office got there first.
The four no-regret bets
If you are deciding where to spend, the field has mostly converged on four moves. Oliver Wyman calls them the no-regret bets: ambient documentation, clinical documentation improvement, coding automation, and electronic prior authorization. These are not moonshots. Ambient docs and CDI are already hitting 90 percent-plus accuracy in specific clinical domains, and coding automation is cutting up to 46 percent off the time on complex cases. When the accuracy is that high and the payback is that fast, hesitating is its own kind of risk.
What HIMSS26 was actually selling
Walk the floor and the theme was impossible to miss. Agentic AI dominated the revenue cycle news at HIMSS26, from Waystar pairing with Google Cloud and Gemini to FinThrive rolling out its Fusion architecture with more than 50 AI and automation use cases. This is the tell that a category has left the lab. When the biggest vendors are shipping agents into the same workflow at once, it is no longer a pilot story. It is a race, and 63 percent of organizations are already in it.
The second-order effects nobody demos
Here is what the vendor booth will not show you. As providers use AI to code and appeal faster, payers use AI to deny faster in return, and the whole thing escalates into an arms race that mostly moves cost around rather than improving anybody care. Then there is the gap: big systems can afford all of this, community providers often cannot, so a divergence in adoption quietly becomes a divergence in performance. And lurking behind every point solution is the EHR vendor, perfectly capable of building an equivalent feature and absorbing your clever startup into a checkbox. None of these show up in a 20-minute demo, and all of them shape whether your investment ages well.
How I would play it
So what do I actually do with all this? Place the no-regret bets, and start with ambient documentation and CDI, because the accuracy and payback are the strongest and the risk is genuinely low. But go in with eyes open. Assume your EHR will eventually ship its own version, so favor tools that integrate cleanly over ones you will have to rip out. Keep half an eye on the community-provider gap if you care about the system, not just your system. And hold on to a little perspective: a faster billing office is a real win, but it is a narrow one. It is not the same as better care, and anyone selling it as such is selling.
- Agentic AI is landing first in the revenue cycle because it is structured, repetitive, and tied to clear dollars, unlike diagnosis.
- About 63 percent of organizations have integrated AI into revenue cycle workflows and 80 percent are exploring generative AI, a 38-point jump in under two years.
- The four no-regret bets are ambient documentation, clinical documentation improvement, coding automation, and electronic prior authorization, hitting 90 percent-plus accuracy in some domains and up to 46 percent less coding time on complex cases.
- HIMSS26 was full of it, from Waystar and Google Cloud Gemini to FinThrive Fusion with more than 50 use cases.
- Watch the second-order effects: a payer-provider AI arms race that shifts cost more than care, a widening gap for community providers, and EHR vendors absorbing point solutions.
Frequently asked
Why is agentic AI succeeding in the revenue cycle first?
Because revenue cycle work is structured and repetitive, has a checkable right answer in whether a claim is paid or denied, carries lower clinical risk than diagnosis, and produces obvious dollar ROI in the current quarter.
What are the no-regret AI bets in the revenue cycle?
Ambient documentation, clinical documentation improvement, coding automation, and electronic prior authorization. These are the priority use cases with the strongest accuracy and payback, achieving 90 percent-plus accuracy in some domains and up to 46 percent less coding time on complex cases.
How widely adopted is AI in the revenue cycle?
About 63 percent of organizations report integrating AI into revenue cycle workflows, and 80 percent are exploring, piloting, or implementing generative AI for it, a 38-point increase in under two years.
What is the payer-provider AI arms race?
As providers use AI to code and appeal faster, payers use AI to deny faster in return. Both sides automate harder, and the net effect often shifts cost between them rather than improving patient care.
What is the disintermediation risk?
Standalone revenue cycle AI tools risk being absorbed by EHR vendors that build equivalent features into their platforms, so a point solution that looks strong today can be commoditized as the EHR ships its own version.
Sources
Naveen Kumar
Healthcare engineering and product executive in Pittsburgh. 15+ years building AI-first patient access, a decade at Treatspace.