Hartford (Conn.) HealthCare saw primary care revenue rise 17% after piloting autonomous coding alongside ambient clinical documentation, according to Kalan Blair, senior director of revenue cycle at the 10-hospital system.
The health system launched the pilot in primary care, pairing ambient listening technology with an autonomous coding platform, according to Ms. Blair. Hartford HealthCare rolled out the technology in stages, beginning with physicians who were more receptive to ambient documentation and gradually expanding as the organization measured results.
“We saw a 17% increase in our primary care revenue, so saw huge gains with this rollout,” Ms. Blair said during a panel at Becker’s 11th Annual Health IT + Digital Health + RCM Conference.
Hartford HealthCare initially audited 100% of the autonomous coding output, with its audit and education team reviewing results and providing feedback. As confidence in the technology grew, the system reduced the level of manual review.
But the financial gains created another challenge: higher patient responsibility.
Ms. Blair said the tool identified charges that previously might not have been captured. For example, a patient arriving for an annual physical could also receive a separately billable sick visit if another condition was addressed.
“Yes, revenue was up, but patient complaints were also up, provider complaints were up,” she said. “How do we make sure that we’re doing this in the best way that we’re serving our patients, our bottom line and our providers?”
The timing amplified the issue. Hartford HealthCare began its pilot in September 2025, but the effect on patients became more noticeable after deductibles reset Jan. 1, 2026.
Ms. Blair said the experience underscored the importance of communication and change management. The system added patient education, signage and physician education after realizing clinicians themselves did not always know the autonomous coding platform was adding charges.
“The biggest thing was the communication and the change management and really making sure that was tight,” she said.
The success in primary care prompted Hartford HealthCare to test autonomous coding in specialty practices, but the results were not uniform.
The system initially expanded into neurology with eight physicians, then halted the pilot after determining that the technology was not a good fit for the complexity of the specialty’s visits and documentation.
“It was not the right fit,” Ms. Blair said. “We weren’t seeing the revenue gains. Our audit and education team was spending way too much time on this when we needed them to now be focused on our next phase.”
Hartford HealthCare then moved into endocrinology and medical weight loss, where Ms. Blair said results have been positive. The specialties offered more standardized documentation and clinicians were receptive to using ambient technology.
The experience has changed how the health system evaluates AI and automation projects. Rather than assuming a tool that succeeds in one area will work everywhere, Hartford HealthCare is building pilots with explicit checkpoints that allow leaders to stop projects that do not deliver the expected results.
“Some are going to work and some aren’t,” Ms. Blair said. “[It’s about] being OK with saying, ‘this wasn’t the right fit, this didn’t work, now let’s pivot and move on.'”
She said creating an environment where teams can experiment — and abandon unsuccessful projects quickly — will be increasingly important as health systems adopt more AI and automation.
“Let’s fail fast, not dwell on the failures and be OK with pivoting, moving on and trying the next thing,” Ms. Blair said. “I think [that] is really going to be key to navigating this new world of AI and automation.”
