Brown Health scales Microsoft Dragon Copilot and AI agents to ease care delivery

*This story was first published in Signal Magazine.

During a busy shift in the emergency department, physician Anthony Napoli listened as an elderly patient described her symptoms—dizziness, heart palpitations, elevated heart rate.

In the past, Napoli would have been thinking through his diagnosis and jotting down details to document the encounter while trying to remain fully present with the patient. The woman was soft-spoken and the conversation took place in a hallway, adding to the challenge.   

That cognitive tug-of-war is now easing with the help of AI tools such as Microsoft Dragon Copilot, which records patient visits and supports care decisions in real time. Napoli used the AI clinical assistant to capture the encounter with the woman, picking up important details he might have otherwise missed amid the noise and bustle of the emergency department.    

“Now, I can focus on what I think is going on with you and hearing you and being with you,” says Napoli, executive vice chair of emergency medicine at Brown University Health in Providence, Rhode Island. “The documentation is happening in parallel with engagement with the patient.”   

That shift reflects a broader transformation underway at Brown Health. Across the rapidly expanding health care system, AI is being deployed not just to automate tasks but to rethink how care is delivered. In a financially constrained environment with growing demand, leaders see AI as a key tool to improve efficiency, expand access and enhance both the patient and clinician experience.   

“We are under constant financial pressure, and we want to meet and exceed expectations for how we deliver high-quality care,” says Adam Landman, MD, senior vice president and chief digital information officer at Brown Health. “We see AI as a key enabler, one part of a broader set of changes needed to transform how we deliver care.”   

A growing challenge: Clinician burnout  

As Rhode Island’s largest health system, Brown Health serves a population of about a million and is a critical care backbone for the region. Since 2024, the organization has undergone significant change, acquiring two Massachusetts hospitals and merging with a large academic physician practice to dramatically expand its size and reach. At the same time, leaders have unified the organization’s technology, rolling out enterprise platforms to manage electronic health records and core operations.   

Layered onto that transformation are pressures familiar across healthcare, from financial strain to labor shortages, an aging population and the growing issue of clinician burnout.   

“One of the main challenges in health care these days is that our clinicians—our doctors, our nurses, our pharmacists, really any of our clinical staff—are increasingly burned out,” Landman says. “There are constant pressures on these team members to see more patients, see them faster and do more with less.”  

“If you look at the studies, burnout rates among physicians in some specialties can be 40% or more,” he continues. “Unfortunately, burnout is then leading to some physicians and others leaving the practice of medicine, which makes the problem even worse.”   

One of the biggest drivers of burnout is documentation. Emergency physicians, who see a high volume of patients with acute needs, carry some of the heaviest documentation burdens, Napoli says. That burden is compounded by a complex health care system in which older, sicker patients increasingly rely on emergency departments for care they can’t access elsewhere. “That creates a really tight environment where you’re trying to care for the patients, but you have to document and reflect the complexity of care that’s going on,” Napoli says.   

As documentation requirements have increased, the technologies designed to digitize and streamline records have not reduced clinicians’ work, and in some cases have made it more difficult. For years, providers have used a mix of technologies and human scribes to manage the load. Each solution offered incremental relief, but none fundamentally changed the equation.   

A few years ago, Napoli began hearing about emerging AI tools for ambient clinical documentation—technology that listens in the background and automatically drafts clinical notes—and saw a potential solution. He soon became a champion of Dragon Copilot at Brown Health, helping lead an effort to make its emergency physician group one of the first fully ambient-enabled clinical departments in the country. Ambient documentation, Landman says, has been “the most profound application of AI so far” for the organization.   

Building an AI strategy  

When Matthew Butler joined Brown Health in May 2025 as its inaugural director of AI and information services, he stepped into an organization that, by his own assessment, was behind. “There really wasn’t an AI strategy,” he says.   

But the urgency was clear. Healthcare is “experiencing a number of different crises,” Butler says, from workforce shortages to access gaps. “Patients can’t get in to see doctors in a timely fashion. Providers are overwhelmed. There’s a massive administrative burden. What AI represents is a great opportunity to empower the people we have to be able to start to meet the challenges head-on.”   

To address that gap, Brown Health established its AI Center of Excellence with a small, specialized team tasked with transforming the health system through AI, from frontline care to back-office operations.   

The center’s first six months were focused on governance—establishing policies, defining acceptable use and building a responsible AI framework. Rather than rushing adoption, Butler says the team prioritized getting buy-in from leaders across the organization. It developed a risk-based approach for evaluating AI tools, with close attention to accuracy, safety and real-world performance.   

One early strategic decision was to focus on tools that integrate directly into existing workflows. Employees were already using Microsoft products extensively, Butler says, making Microsoft 365 Copilot and Dragon Copilot natural starting points. “AI is great, but it’s only great if you can integrate it meaningfully into where people are doing their work,” Butler says. 

Introducing Dragon Copilot 

Microsoft Dragon Copilot became the center’s first major application and a cornerstone of Brown Health’s AI strategy. The AI clinical assistant captures patient conversations (with permission), generates structured clinical notes and delivers them almost instantly. Though it is integrated with clinical systems, the technology is in the background—Landman, for instance, simply starts a session on his phone and sets it on the counter, letting it capture the visit while he focuses on the patient.   

“Once I finish the interaction with the patient, I hit stop recording, and about 20 seconds later, a draft note pops into the electronic health record,” Landman says. “I can quickly review and edit the note as needed, then sign the note, rather than having to finish documentation after hours.”   

An early pilot involving 420 clinicians using Dragon Copilot across multiple specialties produced strong results, Butler says. Clinicians focused more on patients, spent *less time on after-hours documentation and reported a lower cognitive burden.   

The qualitative feedback was also striking. “This is going to extend my career,” one clinician told the team, according to Butler. Others said the technology fundamentally changed how they practiced medicine.   

Brown Health also rolled out Microsoft 365 Copilot across departments. Employees—from executive assistants to analysts—are using it on a daily basis to manage inboxes, summarize meetings and analyze contracts for potential opportunities, Butler says. Others are using advanced Copilot features for data analysis and in-depth research.  

“What we tout across the enterprise is that AI is really about unlocking ingenuity, innovation and creativity, and how each person uses it will be different based on their role and their needs,” Butler says. “What I’m most inspired about is that it brings the fun back into work, with people finding new, novel ways to use it every day.”   

Brown Health’s team and its AI Governance Council used the Researcher agent in Microsoft 365 Copilot while developing a responsible AI policy. The tool enabled them to gather information about best practices at other organizations and rapidly iterate from drafts based on governance council feedback.   

“People were hoping we might have a policy at the end of the first year,” Butler says. “We were able to have a policy in four months that was almost fully ratified. Researcher," he says, “really was a huge enabler.”   

A risk-based approach  

Butler’s team has paired rapid innovation with a deliberately cautious, risk-based approach to deploying new AI tools. Each use case is evaluated for clinical and operational risk and weighed against its potential value before moving forward. “It’s not just about whether a tool works,” says Ali Chambers, a data scientist with the center. “It’s about what happens if it doesn’t.”   

That framework determines how rigorously tools are tested. Lower-risk applications, such as internal knowledge assistants, may progress quickly through prototyping and small pilots. Higher-risk tools, particularly those that could influence clinical decisions, undergo more extensive validation, including deeper testing for accuracy and clearly defined performance metrics.   

In practice, that means most tools move through multiple stages: hands-on prototyping, structured testing with curated question sets, and pilot programs with frontline users to identify edge cases and refine performance. The process relies heavily on close collaboration with clinicians to ensure outputs align with real-world workflows.  

And evaluation continues after deployment. The team tracks adoption and use patterns, gathers feedback and surveys clinicians on trust and perceived value—metrics that can be just as important as technical accuracy for tools that are difficult to benchmark. It also trains employees to engage with AI critically and understand its limitations. “People find it really valuable, but they still don’t trust it, and that’s perfect,” Chambers says. “Because really good use of AI is that you’re a suspicious user.”   

Bring back the joy  

Butler is aiming to position Brown Health as a national—if not global—leader in agentic AI. To get there, he’s focused on a highly collaborative approach, tapping the expertise across the organization to develop solutions that can meaningfully reshape healthcare. “There’s so much potential to unlock innovation internally here at Brown Health,” Butler says. “That’s what I’m most excited about.”   

To turn that potential into practice, leaders are starting to zero in on where AI can have the most immediate impact. Landman sees the greatest near-term opportunity in using AI agents to automate administrative workflows, particularly in areas like revenue cycles and patient access. Over time, he expects AI to become deeply embedded in clinical workflows—augmenting, not replacing, clinicians.   

The ultimate goal, he says, is to expand access to care and give clinicians more time and energy to focus on patients—in other words, to use technology to return medicine to its human roots.  

“It’s going to bring back the joy to the practice of medicine,” Landman says. “Clinicians, including me, went into this because they want to help people. Most did not go into it because they want to interact with computers. And I think we’ll ultimately deliver a better patient experience. Instead of me going off and writing notes for two hours, imagine if I had that time to sit with the patient and talk with them for an extra 15 or 20 minutes. That’s what we’re looking for.” 

*Internal Analysis conducted by Brown Health, Spring 2026.