Building the foundation for agentic AI in healthcare

Grounded in trusted data, embedded in the workflow, and running on always-on infrastructure. Read how three health systems put agents to work.

It's Tuesday morning, and the hospital is already behind in ways it won't recover by the end of the shift. A physician is 10 patients in and hasn't finished a single note, caught in the constant tug-of-war between the patient in front of her and documenting the one she just left. Two floors up, a nurse is several hours into a 12-hour shift and has spent more of it at a workstation than at a bedside. Down the hall, clinical, operations, and finance teams chase the same urgent questions-patient flow, staffing, capacity, and cost-but the answers are scattered across systems, and by the time the data is stitched together, the moment to act has passed. Then an alert reaches IT: the electronic health record (EHR) is slowing down, and the CIO knows every minute clinicians can't reach real-time clinical, pharmacy, and lab data is a minute patient care may be at risk.

These aren't isolated problems. They're the same daily challenges showing up in different forms. Disconnected vendor solutions have added complexity and cost, not insight or care.

Healthcare organizations need solutions that work together to stave off fragmentation.

A connected foundation for agentic AI

Microsoft brings together everything healthcare needs to deploy and scale agentic AI, brought together in one place:

• Microsoft Fabric brings data from siloed systems together into one governed source of truth that agents can reason over.
• Microsoft Dragon Copilot delivers clinical agents that work inside the clinician's workflow.
• Microsoft 365 Copilot and Copilot Studio let anyone across the organization build and run their own enterprise agents.
• Microsoft Azure and the Microsoft security stack provide the resilient, secure infrastructure that keeps agents and care always on.

The result is a shift from fragmentation to coordinated action, with agents that are contextual, knowledgeable, trustworthy, and scalable because they're grounded in Microsoft IQ -your organization's own intelligence-fueled by trusted data and built on enterprise-grade infrastructure.

Because Microsoft IQ is open and model-diverse, customers build on their own IQ and stay in control of it, rather than being locked into someone else's platform. Three health systems show what that looks like in practice.

Valuable agents solve real problems

Brown University Health, Rhode Island's largest health system, serves about a million people and is the critical-care backbone for the region. Since 2024 it has grown fast, acquiring two Massachusetts hospitals and merging with a large academic physician practice-all while navigating financial pressures, labor shortages, and mounting clinician burnout driven largely by documentation burden.

Its first AI application, Dragon Copilot, an AI clinical assistant, addressed that burden directly-letting clinicians across specialties stay focused on the patient instead of the keyboard, with less after-hours documentation and lower cognitive load. One clinician put it plainly: "This is going to extend my career."

It's going to bring back the joy to the practice of medicine. Clinicians went into this because they want to help people. Most did not go into it because they want to interact with computers.

Adam Landman, SVP & Chief Digital Information Officer, Brown University Health

Brown then moved from AI that assists to AI that acts. The team built two emergency department agents embedded in Dragon Copilot-one that answers hospital-specific policy and procedure questions, and one that surfaces on-call schedules and directories so staff can reach the right specialist fast. Those are just two of more than two dozen agents the team has built across the organization-and with Copilot Studio, they went from idea to working prototype in a single day.

The enterprise followed. Brown rolled out Microsoft 365 Copilot organization-wide to their staff to manage inboxes, summarize meetings, and analyze contracts. When leaders needed a responsible-AI governance policy, they put the Researcher agent to work studying best practices and iterating drafts; a policy leaders expected to take a year was ratified in four months.

Trustworthy agents start with trustworthy data

Agents are only as reliable as the data they reason over-which is exactly where Peterborough Regional Health Centre (PRHC) started. It serves a growing, aging population against pressures familiar across healthcare: rising patient volumes, increasing clinical complexity, and hard constraints on staffing, funding, and capacity. Leaders saw the opportunity in AI and recognized it works only when grounded in strong data.

So they started with the data, not the model. The team consolidated 18 production systems onto Microsoft Fabric-one governed source of truth for analytics and AI.

We started from a place of pretty significant data fragmentation and immaturity. If someone needed data, it could take weeks to dig it up, and what came back was often an out-of-date Excel spreadsheet that didn't really answer the question.

Lynn Mikula, CEO, Peterborough Regional Health Centre

The shift was profound: teams moved from waiting weeks for static, out-of-date reports to exploring real-time data together, making faster and more confident decisions across clinical and operational workflows. Mikula's lesson reflects a core principle of an agentic strategy-scalable AI doesn't start with models-it starts with data. With a governed foundation in place, PRHC created the conditions for agents to scale on data they can trust.

And the payoff is showing up where it matters most-in patient care. PRHC attributes improvements across several emergency departments and inpatient flow metrics from January to March 2026 to this operational effort. Fewer patients waited for inpatient beds at 8 AM, dropping from 34 to 19, and the wait time for an inpatient bed fell 43%, from 56.8 hours to 32. As access improved, the share of patients leaving without being seen decreased from 10% to 8.8%, while overall experience scores rose from 77% to 80%. A new hospitalist swing shift improved consult response times by about 20% while helping balance inpatient workload and support care quality. Together, these changes contributed to longer-term gains in patient flow, including a roughly 70% reduction in ambulance offload times.

Reliable agents run on resilient infrastructure

Agents that clinicians depend on can't go dark. That made infrastructure the priority at Franciscan Health, which migrated its Epic EHR to Azure for performance, scalability, disaster recovery, and security on a future-proof platform. The stakes were absolute. "If Epic is down, we don't have access to clinical data. We don't have access to pharmacy; we don't have access to lab results," says CIO Charles Wagner. "So, from the minute that's down, patient care is at risk." The cost is just as stark: "it costs us $10 to $12 million a day every day we're down."

The cutover was the first proof point. "Once we did the cutover, I don't think we got a single call. The clinicians and staff didn't even know it happened," Wagner says. "That's how we define success." Then the results compounded across performance, cost, and recovery-Franciscan can now bring its disaster-recovery environment live in under an hour, down from the 8 to 12 hours it once took, keeping continuity of care intact even in a major incident.

Since we migrated to Azure, we can do a failover and be up in 30 minutes. This is something we would never be able to do in our previous environment.

Charles Wagner, CIO, Franciscan Health

Franciscan's Azure migration delivered:

• $45M in savings over five years.
• 33% lower infrastructure costs.
• 50% faster application response.

And because Epic builds on Microsoft technology, Franciscan gained a single-vendor advantage that makes the next agent, the next model, the next workload easier to bring online-the difference between fighting your infrastructure and building on it.

One platform. AI you can trust

Three health systems, three starting points, one pattern. The value isn't in isolated tools: it's in a connected platform where agents are grounded, embedded, and reliable.

• Brown Health put clinical and enterprise agents to work across two dozen departments.
• Peterborough Regional Health Centre built the trusted data foundation agents reason over.
• Franciscan Health established the resilient infrastructure that keeps agents and care running.

Everything works together, which makes it easier to bring new solutions online and get more value from our investments.

Charles Wagner, CIO, Franciscan Health

That's the shift: not more tools, but a system that works as one, with data, workflows, and intelligence that understands context, your business processes, and your teams' interactions, allowing agents to move from signal to action. Fabric makes the organization AI-ready. Dragon Copilot transforms the clinical workflow. Copilot connects teams, decisions, and execution. Copilot Studio lets anyone build scalable, repeatable agentic workflows. Together, these capabilities are grounded in Microsoft IQ and run on Azure, providing the resilience, security, and scale healthcare organizations depend on. From the analyst using Fabric, to the clinician utilizing Dragon Copilot, to the rest of your employees using Copilot, to the builders creating in Copilot Studio, the organization finally runs on the same intelligence, and the same agents, turning fragmented signals into coordinated action, and data into better patient care.

So, when the next shift begins, clinicians can spend more time caring for patients, leaders can make decisions with confidence, and organizations can act on insight before the moment to act has passed.