Microsoft Expands Dragon Copilot to Optimize Radiology Workflow

Seamless human + AI collaboration: The next frontier in radiology tide of images, data and documentation—each with varying degrees of complexity and urgency. And every moment a radiologist spends wrestling with fragmented technology is a missed opportunity to make a difference when it matters most.

This challenge extends into the AI landscape as fragmentation occurs when tools address narrow problems but create extra steps and poor integration into broader clinical workflows. Yet, hope is on the horizon. Microsoft is breaking down barriers between human expertise and AI to usher in a new era. Clinicians and AI can work side-by-side, allowing clinicians to get faster answers, deeper insights, and support better patient care.

Microsoft is empowering health care organizations to advance their enterprise AI strategies and elevate workforce productivity. “We’re taking the concept of frontier firms and applying it specifically to health care organizations,” said Sheela Agarwal, MD, PhD, MBA, chief medical information officer for diagnostic imaging and AI at Microsoft. “A frontier health care organization embraces multi-agent systems that work alongside clinicians, instantly delivering trusted information and clinical insights while orchestrating and automating tasks.”

Dr. Agarwal cited Stanford University’s successful use of Microsoft’s multi-agent orchestration to coordinate preparation for a tumor board. Stanford built multiple specialized agents that pulled together electronic health record and imaging data, guidelines, literature and current trial information in real-time, allowing the team to make clinical decisions within days instead of weeks.

“We want the same shift to occur in radiology with Dragon Copilot so radiologists can benefit from human + AI collaboration without disrupting their existing reporting workflow,” Dr. Agarwal said.

Expanding Dragon Copilot for Radiology

Microsoft is expanding its established AI clinical assistant, Dragon Copilot, by introducing capabilities tailored to radiology. Serving as a companion for PowerScribe One, Microsoft’s market-leading reporting solution, this combination delivers a unified workflow that streamlines reporting, surfaces information and automates tasks.

Dragon Copilot was launched in spring 2025 to reduce administrative and cognitive burdens for clinicians by helping them with clinical documentation, order capture and care coordination. Originally designed to support physicians at the point-of-care, Dragon Copilot has since expanded to support nurses and now radiologists.

It also connects partner and self-developed AI applications and agents directly into the end-to-end clinical workflow, helping teams work more efficiently while maintaining consistency and compliance across the organization. 

Generative, Multi-modal and Agentic AI

For radiologists, Dragon Copilot works with PowerScribe One to deliver generative, multi-modal and agentic AI in a secure, scalable and extensible way. “It’s our way of bringing cloud-native features and AI directly into the radiologists’ workflow,” Dr. Agarwal said.

By integrating with PowerScribe One, Dragon Copilot can further streamline report creation without interrupting the radiologist’s interpretation. Now research-grade, multi-modal AI models—whether developed by the organization or a third party—can generate draft report content through automatic image analysis.

Dragon Copilot can further optimize reports by integrating AI results from an ecosystem of third parties. Using a partner model, Dragon Copilot can alert the radiologist about missing billing information in their draft report, avoiding the need to issue an addendum later.

Additionally, a chat experience gives radiologists quick access to credible information with trusted sources from the internet and large language models without requiring them to open additional applications and windows. “We see chat as a secure, natural place where radiologists will eventually interact with additional AI apps and agents that understand the full patient context and are connected to other health care IT systems,” Dr. Agarwal said.

The prior report summary—Dr. Agarwal’s favorite capability—distills a patient’s relevant prior radiology reports into concise bullet points. “Before looking at a case, you’ll know what brought the patient in, their history, a timeline of their disease process, and other relevant clinical findings regardless of the anatomical region being studied,” she said. “As an abdominal radiologist, I always look at the most recent abdominal CT, but I don’t necessarily have time to check brain or chest imaging that could be relevant for my interpretation.”

For both chat and prior report summarization, Microsoft has safeguards and guardrails to help ensure information provided is accurate and comprehensive. “Dragon Copilot provides provenance and links to pertinent facts which the radiologist can hover over and quickly verify,” Dr. Agarwal said.

Collaborating for Real-world Impact

Microsoft works closely with radiologists, including teams from the University of Rochester (UR) in New York, to elicit valuable feedback and to stress-test Dragon Copilot capabilities. “Microsoft stands out as a partner that understands the radiology workflow and knows how to integrate AI in a way that truly addresses our daily pain points,” said Sean Cleary, MD, vice chair of informatics for imaging sciences at UR.

From administrative burdens to staffing shortages, challenges are only intensifying, according to Dr. Agarwal. “There’s so much red tape and manual, repetitive tasks keeping us from doing clinical interpretation while shortages and departmental cuts mean more tasks are transferred to radiologists,” she said. “What makes all of this so exciting for me is how it will help ease cognitive burden and eliminate repetitive tasks so radiologists can work at the top of their license.”

Learn more about Dragon Copilot by visiting the Microsoft Booth 1311, South Hall A, or go to aka.ms/DragonCopilot.