Most healthcare organizations are already using AI in some form or another. What they’re still trying to figure out is whether those tools are actually helping clinicians and care teams do their jobs better, and whether they're actually moving the needle in value-based care.
In a recent webinar with AMGA, Navina’s recent webinar with AMGA, Achieving the Real ROI of AI: Optimizing Clinical, Financial and VBC Success for 2026, leaders from primary care organizations talked candidly about what’s working, what isn’t, and how they’re thinking about ROI as AI use matures. Dana McCalley, VP of Value-Based Care at Navina, was joined by Dr. Flora Zarcu-Power, Director of Primary Care at PriMed Northeast Medical Group, and Dr. Eric Penniman, Executive Medical Director at Summit Medical Group.
Below are five key takeaways from the conversation.
The ROI from AI cannot be financial alone
Narrow definitions of ROI rarely capture AI’s full potential value.
“In 2026, my focus is pragmatic,” said Dr. Zarcu-Power. “Choosing AI solutions that solve real pain problems, define clear metrics, and ensure that the value we bring is both human and financial.”
While financial performance remains critical, the speakers emphasized that AI’s impact often appears first elsewhere: improvements in clinician experience, reduced burnout, and better patient engagement.
“Financial savings are certainly very important,” Dr. Zarcu-Power added, “but equally critical are reductions in clinician burnout, improved patient satisfaction, and stronger workforce retention.”
This broader view shifts the lens through which AI success should be evaluated. A solution that meaningfully improves clinician experience or prevents attrition may ultimately deliver more durable value than one that produces short-term cost savings alone.
Clinician adoption depends on seamless workflow fit
AI Adoption is driven less by enthusiasm for AI and more by whether tools make daily work easier.
“For people to adopt it, it can’t be more clicks,” said Dr. Penniman. “It has to be easier to get the job done.”
At Summit Medical Group, tools that reduced friction gained traction quickly. Dr. Penniman described how ambient listening spread rapidly because clinicians felt the benefit immediately—saving time and reducing after-hours documentation.
Tools that sit outside existing workflows or require clinicians to toggle between systems struggle to gain trust. By contrast, solutions that integrate directly into the point of care, without adding administrative or cognitive burden, see faster and more durable adoption.
As Dr. Zarcu-Power noted, clinicians don’t want another system to manage. They want support that fits naturally into how care is already delivered.
Early signals matter more than delayed ROI data
In value-based models, financial outcomes often lag behind operational change. Relying solely on claims-based data can delay decisions and stall momentum.
“At Summit, we rely on claims-based data for Medicare Advantage and ACO Reach,” Dr. Penniman explained. “That data is always delayed. Come June, we’ll have a clearer picture.”
In the meantime, Summit tracks early indicators such as provider satisfaction, time saved, efficiency gains, care gap closure, and RAF performance. These signals help determine whether AI initiatives are delivering value and whether they warrant scaling.
Dr. Zarcu-Power emphasized the importance of defining success upfront. “We’ve seen pilot projects that did not have clear evaluation frameworks, and they struggle to justify scaling. In 2026, organizations need to define success up front and measure against it.”
The goal isn’t to lower the bar for ROI, but to measure the right indicators at the right stage.
Context is king, especially in value-based care
In value-based care, a single visit tells only part of the story. Clinicians need longitudinal context to manage risk, close care gaps, and make informed decisions.
“One of the frustrations for most primary care providers is the disparate EHRs out there,” said Dr. Penniman. “You often don’t have the information you need when your patient is right in front of you.”
Summit Medical Group operates across multiple hospital systems and EHRs. Without aggregation and synthesis, clinicians are left to manually reconstruct patient histories. Dr. Penniman described how unified patient summaries change that experience.
“Now all of a sudden, we’re getting documents pulled in and teed up for us. The AI has already digested that information and identified what’s important,” he said.
Dana McCalley shared a real example of what happens when context is missing: a patient with a prior heart failure hospitalization who was seen multiple times in primary care without the care team knowing about the event. The result was a missed opportunity for intervention and higher downstream costs.
In value-based settings, connecting data across encounters and systems is essential for both quality and financial performance.
AI adoption is a change management initiative, not a one-time rollout
Governance is often underestimated in AI adoption. “AI is only as good as the data we train it on,” Dr. Zarcu-Power said. “Healthcare data is notoriously fragmented and unstructured.”
She cautioned against what she described as “black box liability,” where vendors cannot clearly explain how recommendations are generated. “Physicians need to understand where the answer is coming from,” she said.
Dr. Zarcu-Power advocated for greater transparency through what she referred to as “AI nutrition labels,” including model cards that outline performance, indications, limitations, and how models perform across different subpopulations.
Workforce readiness is equally critical. Clinicians and care teams must be trained not only to use AI, but to critically evaluate its outputs. Without guardrails, AI can amplify poor habits as easily as good ones.
Dr. Penniman echoed this point, explaining that Summit implemented review processes to ensure new diagnoses were clinically supported. He noted that “AI will amplify those doctors who aren’t as diligent, unless you put the right guardrails in place.”
As Dr. Zarcu-Power emphasized, AI is not a one-time install. “It requires monitoring, recalibration, and continuous governance. It’s a living system that requires stewardship.”
Looking ahead to 2026
As organizations plan for the year ahead, several themes stood out:
- Define ROI beyond cost savings
- Choose tools that fit clinician workflows
- Track early operational signals
- Prioritize longitudinal patient context
- Treat AI adoption as sustained change management
As Dr. Zarcu-Power concluded, “the organizations that will thrive will be those that prove AI value not only on the balance sheet, but to the people who deliver care and receive care.”
Watch the full webinar to hear the complete discussion.
.png)
.png)
