At Becker’s Health IT + Digital Health + RCM Annual Meeting, healthcare leaders from KeyCare, Codman Square Health Center, and Nabla came together to discuss a question that is becoming increasingly important as ambient AI matures: What happens downstream when clinical documentation gets better?
The panel, “Revenue Starts in the Exam Room: How 2 Health Systems Are Improving Documentation, Coding and Financial Performance,” brought together Corey Ogden, MD, Senior Medical Director of Product and Medical Informatics at KeyCare; Renee Crichlow, MD, former Chief Medical Officer at Codman Square Health Center; and Ed Lee, MD, Chief Medical Officer at Nabla.
Across two very different care models, one theme was consistent: documentation, coding, clinician capacity, and revenue cycle performance are deeply connected. Improving one requires looking at the system as a whole.
Accurate Coding Starts With Consistent Documentation
For KeyCare, a national virtual care platform, coding accuracy is particularly important because of its range of payer agreements. While clinicians spend only a few seconds selecting a code after an encounter, the organization historically invested significant manual effort downstream to ensure the code aligned with the documentation.
Ambient AI offered an opportunity to address the problem earlier. By creating more consistent documentation of what happened during the encounter, KeyCare could more reliably connect the clinical note to the appropriate code.
KeyCare piloted Nabla’s E/M coding support with clinicians across a range of baseline coding performance.
Among clinicians whose coding more frequently required review, review rates fell by 15 to 20 percentage points during the initial few weeks.
Importantly, clinicians who were already coding accurately maintained that performance, giving the team confidence to expand the capability more broadly.
The early use also reinforced an important implementation lesson: accuracy alone is not enough. When the coding recommendation initially required clinicians to navigate to another app, KeyCare provided feedback that the additional step would create too much friction. Nabla subsequently brought the recommendation directly into the note workflow.
As Ogden explained,
“the goal is not to ask clinicians to move faster. It is to design the encounter so that efficiency happens naturally while maintaining quality.”
Clinician Efficiency Is an Access Strategy
At Codman Square, the challenge looked different, but the connection between documentation and organizational performance was just as clear.
Crichlow described technology adoption through the lens of a resource healthcare organizations cannot buy: time. For a community health center facing significant primary care workforce and access challenges, reducing administrative work was not simply a clinician experience initiative. It was a way to create more sustainable clinical capacity.
That philosophy shaped how Codman Square approached ambient AI. Rather than using time savings to simply ask physicians to squeeze additional patients into already demanding schedules, the organization focused on creating a more sustainable system that could help clinicians maintain or increase their patient-facing hours over time.
Crichlow summarized the connection simply: even one less click can contribute to patient access when multiplied across clinicians and encounters. When clinicians can spend less time working with the EHR and more time working with patients, those efficiencies ultimately create more patient-facing capacity.
Revenue Cycle Transformation Is a Team Sport
Perhaps the clearest takeaway from the discussion was that technology alone does not create these outcomes.
At Codman Square, introducing ambient AI became an opportunity to rethink the broader system around documentation and revenue cycle performance. Clinical and revenue cycle leaders began meeting regularly, reviewing denials and cycle times, and creating tighter feedback loops between frontline clinicians and the teams responsible for downstream processes.
KeyCare took a similar approach to implementation. Ogden emphasized the importance of testing interventions, measuring what actually happens, gathering clinician feedback, and being willing to adjust when the original hypothesis is wrong.
Both experiences point to the same lesson: AI implementation cannot be “set it and forget it.” Successful adoption requires engagement, transparency, continuous feedback, and trust between clinicians, operational leaders, and technology partners.
As health systems move beyond the first generation of ambient AI, the opportunity is becoming broader than documentation alone. Better documentation can create a stronger foundation for more accurate coding, more efficient revenue cycle processes, greater clinician capacity, and ultimately better patient access.
The exam room may be where the encounter begins, but what happens there shapes everything that follows.
