At the AMGA Annual Conference, Navina hosted a panel discussion around a question many healthcare leaders are quietly wrestling with: if AI is so widely used, why isn't it actually reducing clinicians’ workload?
An audience poll captured the disconnect immediately. About 80% of physicians reported using AI in some form, yet 44% said it hasn't meaningfully reduced their workload. When asked to describe their biggest challenge in one word, one theme stood out: adoption.
The takeaway was clear. AI use isn’t the problem. Integrating it into clinical workflows is.
Workflow friction kills AI adoption
The panelists were direct: the biggest barrier to AI isn’t performance, it’s usability in real clinical environments.
Dr. Bhavana Vora, Family Physician & Board Chair at Summit Medical Group, shared details of an early AI pilot at Summit that looked promising. The tool surfaced relevant insurance and HIE data directly into documentation, exactly what physicians had asked for. In practice, it added 15 extra clicks.
“We did not engage our physician providers in the initial decision-making, which was a big failure on our part,” she said. “Adoption is probably my biggest challenge in clinical workflow.”
Even small amounts of friction can derail adoption. One extra screen or login is often enough to deter engagement.
“It almost has to be like an easy button,” said Dr. Vora. “If it cannot be just right smack in front of my face, it’s not going to get used.”
As Navina has consistently seen, the tools that earn adoption are the ones embedded directly in the EHR, not bolted on alongside it.
AI governance can’t be reactive
Dr. Michael Millie, Chief Medical Officer at HarmonyCares, described a different facet of the adoption challenge: governance. He initially believed his organization wasn’t using much AI. In reality, clinicians were already experimenting with tools independently, but without oversight or organizational alignment.
“Governance is the big mistake,” he said. “It’s happening. Your providers are using it. Get in front of it and develop a process so you can govern it.”
Adoption is already happening, often faster than organizations realize, says Millie. The challenge for leadership is ensuring that AI is introduced within a framework for quality, safety, and consistency rather than spreading informally across teams and workflows.
Organizations that wait to establish governance until after tools are widely used will find themselves chasing a runaway train.
Clinicians trust AI when it brings value
For both organizations, ambient documentation became an important proving ground for AI adoption. It addressed two of the biggest challenges discussed throughout the session: clinician trust and organizational alignment.
At Summit Medical Group, physicians using ambient listening spent about 30% less time on documentation and were able to see additional patients.
The impact went beyond efficiency. Ambient documentation changed perception. Skeptics became advocates. Once clinicians trusted AI with something as sensitive as their notes, they became more open to additional use cases like risk adjustment and care gap management.
“Ambient listening is the world’s best way to sell this to your providers,” said Dr. Millie.
For leadership teams, tools like ambient documentation also create a more manageable starting point for governance.
Rather than reacting to scattered AI adoption across workflows, organizations can introduce AI through a clearly defined use case with measurable outcomes and clinician buy-in.
The pattern is clear: safe and governable AI adoption accelerates when AI delivers immediate, visible value with minimal disruption to the clinical workflow.
Incentives drive AI adoption
Both organizations also stressed that clinicians are far more likely to adopt AI when incentives are aligned with its use.
At HarmonyCares, clinician compensation is tied to outcomes, with AI tools supporting performance along the way. At Summit, shared savings flow directly back to physicians, with clear expectations built into the model.
"If your compensation is dependent on it, everybody buys in much easier," said Dr. Vora.
Measure AI impact before formal rollout
Another gap surfaced clearly during the session: while most organizations have begun using AI, very few have implemented ways to measure its impact.
Dr. Millie cautioned against adopting multiple tools at once without clear evaluation criteria. "You can go buy 23 things that are all supposed to reduce readmissions, and you'll have no idea which one is working."
At Summit, new tools are piloted for at least six months before formal rollout with clear KPIs defined before scaling. Dr. Vora offered a concrete example of what rigor looks like in practice: in value-based care, even a 0.1 improvement in RAF score has a calculable dollar value. That’s the level of measurable clarity an organization should establish before committing to any solution at scale.
AI only works when clinical workflows change
AI is already widespread in clinical settings. What separates the organizations seeing real impact isn’t whether they’ve adopted it. It’s whether they’ve structured their data, governance, incentives, and measurement around it.
Until AI fits seamlessly into the way clinicians work, its promise to reduce workload will remain largely unrealized.
Click here to watch the full recording of this session.
.png)
.png)
