Skip to content
Sammie Kim

Dragon Copilot for Radiology

Designing AI-assisted reporting that integrates into radiologists’ existing workflows, from the long-term product vision to the first release, Copilot chat.

Role
Product Designer
Duration
2025–2026
Team
UX, UXR, PM, Eng, RAI, Accessibility
Focus
Design concept strategy, storyboarding, interaction design

01 — Problem

Growing workloads, fragmented workflows

Radiologists face growing workloads while frequently searching across reports and external resources to understand complex findings. These fragmented workflows create interruptions and add time to reporting, revealing an opporunity for AI to bring relevant information and guidance closer to the point of work.

Illustration of a radiologist dictating a report in PowerScribe One, with PACS imaging and the patient chart open alongside it.
Studies read per radiologist, per day
100–150
Radiologists interviewed and observed
13
  • “We are all up to our ears at work and it doesn’t seem like this increase in workload is going away.”

    — Radiologist

02 — Public Preview

AI assistance within the reporting workflow

Private preview

Dragon Copilot introduces AI assistance directly within the PowerScribe One reporting experience. The companion helps radiologists access information and work through questions without leaving their workflow, reducing context switching while keeping them in control.

Early feedback from radiologists

  • “Helps to make you more accurate. It prompts you with context you may have otherwise ignored.”

  • “I can look at one screen, get a general idea of the reports for relevant priors.”

  • “It tells me all of their imaging findings in a more concise way and gives me context comparing it to the prior.”

03 — Design overview

Inside the Copilot experience

A closer look at how Copilot supports each step of reporting, and the lightweight feedback loop built into the experience so it keeps improving through the private preview.

Ask a clinical question and get an answer with cited references, without leaving the reporting context.
Review prior reports summary to understand patient's history at a glance. Provenance and citations are surfaced for each finding.
Draft a report, generate an impression, and receive billing recommendation.

Feedback interaction

A lightweight way to rate the product and flag specific AI features, which drove a 300% increase in feedback submissions.

The feedback window can be placed anywhere in the workspace, so radiologists can give input without losing their place in the report.
Rate the product and individual AI features, leave a comment, and submit.

Full case study, including design decisions and process, available on request. sammiekim1227@gmail.com