Accountable Care Leaders Spotlight Next Phase of AI at NAACOS 2026 Spring Meeting

Experts wrapped up the National Association of Accountable Care Organizations (NAACOS) Spring 2026 Meeting in Baltimore, Maryland, last week by highlighting the growing shift of artificial intelligence (AI) from experimental to operational across clinical workflows.

CMS Leader Details Health Tech Ecosystem Strategy in Fireside Chat

Following the lunch break on the conference’s final day, Aisha Pittman, MPH, senior vice president (VP) of government affairs at NAACOS, led a fireside chat with Alberto Colon Viera, chief health technology officer at CMS.¹

During their discussion, Colon Viera positioned CMS as a central force in health care transformation due to its scale, policy reach, and convening role. He highlighted increased cross-agency collaboration, particularly with the FDA, as key to accelerating innovation. Although CMS remains a complex organization, Colon Viera emphasized ongoing efforts to modernize its infrastructure and improve data exchange across the system.

A key focus is a voluntary CMS-led health tech ecosystem initiative. He explained that it originated from a formal request for information process launched last summer, which surfaced persistent interoperability challenges.

The initiative seeks to give patients access to longitudinal health records through apps of their choice, supported by standardized identity verification. The goal is to reduce reliance on fragmented patient portals while enabling providers to access timely data at the point of care. Payers are included, with participation tied to sharing claims data when patient consent is provided.

Colon Viera noted that the framework introduces “CMS-aligned networks,” which connect health information exchanges and other data systems to enable secure, identity-verified data exchange across patients, providers, and payers. He emphasized that participation is voluntary, with collaboration occurring through working groups and early adopter testing.

A key component of these networks is patient-facing apps organized into 3 categories: tools that replace manual processes, conversational AI applications, and chronic condition management platforms. These apps connect through CMS-aligned networks to enable more seamless data exchange at the point of care to reduce administrative burden and improve access to information for both patients and clinicians.

CMS is also addressing persistent data fragmentation through a phased interoperability strategy. While leveraging existing Consolidated Clinical Document Architecture–based systems in the short term, the agency is moving toward a Fast Healthcare Interoperability Resources–based framework to improve data normalization, patient matching, and aggregation.

Colon Viera added that CMS is aligning these technical efforts with financial incentives through the ACCESS model, a value-based care initiative that connects data interoperability with payment reform. The model integrates technology-enabled care across conditions such as cardiometabolic, musculoskeletal, and behavioral health, reinforcing a performance-based structure tied to quality metrics. To conclude, he encouraged audience participation.

“Go to CMS.gov health tech ecosystem…we're looking for providers that want to join; just pledge to be there,” he said. “There's a contact form that you can complete to indicate your interest, and pledge [to be] a member of the health tech ecosystem. Join the ecosystem, be at the table, help us solve these technical problems, and let's make it a reality for everyone.”

Closing Plenary Highlights Real-World AI Impact

To round out the conference, experts participated in the closing plenary, “AI in ACTION: New Solutions that Advance Accountable Care,” moderated by Brian McDonough, MD, chairman of the graduate medical education committee for the department of family medicine at St. Francis Hospital.²

The panel consisted of James Barr, MD, of Atlantic Health; Lyell Jones, MD, of Mayo Clinic; Tim Puri, MBA, of Curana Health; Brian Sikora, MPA, of Duly Health and Care; and Tina-Ann Kerr Thompson, MD, MBA, of Emory University.

Each speaker shared a recent AI initiative from their organization, starting with Barr. Atlantic Health turned to AI to address persistent challenges, including uncontrolled diabetes and hypertension, as well as clinician burnout. It deployed an AI-powered chatbot app that functioned as a 24/7 digital assistant nurse, increasing touchpoints between visits. Powered by approximately 5000 evidence-based guidelines, the tool integrated electronic health record (EHR) data to flag medication changes, assess prescription adherence, monitor for adverse effects, and coach patients on diet and exercise.

Within 6 to 12 months, 64% of patients with previously uncontrollable diabetes achieved control, with an average hemoglobin A_1c reduction of 3.66 points. Patients with hypertension saw an average systolic blood pressure drop of 34 points. Barr noted that the chatbot is a valuable new data stream, as patients often disclose more to it than to clinicians.

“You have your claims data, pharmacy data, lab data, and EHR data, [but] you need to go to the source, the patient,” he said. “This is where AI can bring the data in. All of a sudden, together, [we] as the humans and AI as a data source, we start to write the next chapter in the health of our patients.”

Next, Jones highlighted one of Mayo Clinic’s AI initiatives, “Colon Pilot.” He explained that colonoscopy results are often given out of context of the patients’ broader history and risk. Therefore, it is not always clear when the next screening should occur.

To solve this problem, Colon Pilot uses a large language model to summarize the colonoscopy report and compare the findings to evidence-based guidelines. It also pulls relevant patient history from the EHR to assess forward-looking risk before generating a standardized, readable recommendation for timing of the next screening.

Jones noted that the tool ensures consistent, guideline-based screening intervals while minimizing manual chart review and interpretation for clinicians. He added that Colon Pilot has been validated, with success seen in practice, originating within Mayo Clinic’s community health network.

“This was not a project that was initiated in a large academic center and then disseminated,” Jones said. “This is something that I think shows the accessibility and the democratization that this kind of technology brings.”

After, Thompson detailed the benefits of AI-driven outbound calls, as validated by Emory University. She explained that the institution was underperforming on hypertension quality measures and had no staff capacity left during a “4th quarter sprint.” To help with this, AI agents called patients, asked them if they had an appropriate BP cuff, guided them through proper measurement, and collected spoken BP readings and other information.

Thompson reported that about 90% of patients who completed the calls said the experience was good or better. The tool also helped Emory University improve from 1 to 4 stars on the payer’s hypertension metric, where it was used. At the same time, staff were happier and felt supported by the AI.

“It showed our employees that we were willing to launch something to help them,” she said. “They were never worried about their jobs being replaced…They were thrilled to have another tool in their toolbox.”

At Duly Health and Care, Sikora and his colleagues wanted to prevent avoidable hospital admissions. Consequently, they built a prediction model to flag patients with high 6-month admission or readmission risk. He explained that approximately the top 8% of scores captured most admissions, so the organization created a nurse care management program focused on that top 8%.

Among enrolled patients, the model helped reduce acute admissions by 40%, emergency department utilization by 30%, and overall medical costs by 20%. Based on these results, Sikora highlighted that AI can work behind the scenes while human care teams do the visible work.

“Your AI does not need to be visible in order to be valuable,” he said. “If you look at it from a workflow perspective of what you are trying to achieve and trying to match the right solution to the right step of a process, you can have results like this.”

Lastly, Puri described an AI used to prioritize patients in senior living communities. He explained that Curana providers go into senior living communities and must decide which patients to see and in what order. Traditionally, this prioritization was manual, based on reactive “chart hunting” and influenced by factors like clinical risk and recent events.

To help with this, the organization built a predicted schedule plus a previsit summary that pulls data from about 8 sources, ranks patients to see at each facility, and generates a suggested task list for each visit. Puri described early outcomes, noting that there have been more encounters per day, especially with patients who have open gaps or important work to be done, with providers reporting satisfaction.

“Providers are saying it saved them 30 to 60 minutes a week, which is fantastic,” he said. “But really, [the fact] that they're doing more visits is the really good outcome there, and providers are really happy with the tool to date, because it solves a problem for them.”

Challenges, Opportunities in Scaling AI Across Health Care

The experts’ experiences with these tools helped to shape the subsequent discussion. Although AI is being rapidly adopted across the health care landscape, implementation has not been without significant friction.

Barr emphasized persistent clinician resistance at Atlantic Health, particularly among physicians already overwhelmed by administrative burden. Thompson added that skepticism often stems from unfounded assumptions, such as older patients being unlikely to use digital tools, despite data suggesting otherwise.

Panelists agreed that overcoming resistance requires reframing AI as a patient-centered solution rather than a cost or operational initiative. Barr said he positions AI tools around improving outcomes for patients with uncontrolled diabetes or hypertension and underscored a “help me understand you” approach, which engages clinicians directly to align AI tools with real-world workflows. Thompson added that organizations should challenge assumptions and pilot solutions rather than overanalyze them; when AI demonstrably reduces burden and improves care, adoption tends to follow naturally.

Reducing burnout emerged as a central value of AI, enabling more frequent patient “touches” through chatbots, automated outreach, and virtual monitoring, all without adding to clinician workload. Barr shared that in one case, only about 6% of patient-initiated messages reached a physician’s inbox, with the rest handled by AI-supported education and triage.

These efficiencies translate into both measurable gains, such as reduced inbox volume, and qualitative improvements, including lower stress and better work-life balance. Jones pointed to the rapid uptake of ambient listening and AI scribes as further evidence that clinicians “vote with their feet” when tools meaningfully reduce documentation burden.

As adoption grows, governance and equity have become increasingly important. Thompson described structured oversight processes to evaluate security, appropriateness, and potential bias, including dedicated monitoring of interactions with underrepresented populations. However, she cautioned against holding AI to a higher standard than human decision-making, noting that human bias is often less visible but equally impactful. In addition, Sikora advocated for practical governance frameworks, including maintaining a clear inventory of where and how AI is used in workflows, both to ensure compliance and to support organizational learning.

Data fragmentation remains a parallel challenge. Puri highlighted the complexity of aggregating data across facilities, EHRs, and external sources. However, once unified, he noted that AI can move beyond summarization to deliver actionable insights, identifying care gaps, risks, and intervention opportunities. Puri also described AI-driven medication review tools that assess drug interactions, deprescribing opportunities, and guideline adherence at scale, as well as experimental use of AI connected to medical literature databases for real-time, evidence-based clinical queries.

Looking ahead, panelists expect AI to accelerate the shift from episodic to continuous care. Remote monitoring, chatbots, and real-time data streams allow clinicians to track patient status between visits and intervene earlier, better aligning care delivery with the continuous nature of chronic disease.

Despite the momentum, panelists stressed that AI’s role is to augment, not replace, clinicians. As Barr put it, the human clinician remains the “secret sauce.” When implemented effectively, AI strengthens that patient-clinician relationship by reducing burden, improving access, and enabling more proactive, personalized care.

“Don't let AI do its thing and run away with this,” he said. ”Bring you as the human to the AI, or bring the AI to you to make it feed what you're doing. You are doing what the patient really needs. They need someone to trust and respect in this health care world.”

References

1. Colon Viera A, Pittman A. Fireside chat with Alberto Colon Viera, chief health technology officer, CMS, and Aisha Pittman, senior vice president, government affairs, NAACOS. Presented at: NAACOS Spring 2026 Conference; April 22-24, 2026; Baltimore, MD.
2. McDonough B, Barr J, Jones L, Puri T, Sikora B, Kerr Thompson TA. AI in ACTION: new solutions that advance accountable care. Presented at: NAACOS Spring 2026 Conference; April 22-24, 2026; Baltimore, MD.