From Dictation to Generation: How AI is Redefining Reporting

Generative AI foundation models are emerging at a moment when healthcare faces growing workforce shortages, rising imaging volumes, and pressure to improve efficiency without compromising quality or safety. As these models begin producing clinically relevant draft findings, practices must rethink reporting workflows, infrastructure, and oversight to ensure responsible adoption. Attendees will gain essential context on why AI-assisted report drafting is gaining momentum now, and why successful deployment depends as much on workflow design, governance, and human factors as on model performance.

Participants will learn practical lessons from early real-world deployments of multimodal foundation models used for clinician report drafting. The session will explore key implementation decisions, including whether AI should generate full draft reports or targeted findings, how text outputs can be linked to imaging evidence, and how to integrate these tools into radiologist workflows with minimal friction. Attendees will examine tradeoffs between efficiency and safety, the implications of standardizing report language versus preserving individual radiologist style, and how radiologists define acceptable performance thresholds before trusting AI in daily clinical use.

Finally, attendees will take away a clear framework for building post-deployment quality and safety programs. Topics include automated performance monitoring, human-in-the-loop review strategies, tracking clinicians edits and engagement, and mitigating risks such as bias, hallucinations, and over-reliance on automation. The session is designed to be highly interactive, combining short focused presentations with audience polling, real-world “what would you do?” scenarios, shared institutional experiences, and a moderated discussion that encourages participants to actively reflect on how generative AI should be implemented to support clinical excellence and patient care.

Objectives

• Describe the clinical, workflow, and infrastructure considerations required to implement generative AI foundation models for clinician report drafting.
• Differentiate between approaches to AI-generated reporting outputs, including full draft reports versus targeted findings, and evaluate their implications for efficiency, safety, and radiologist oversight.
• Implement key elements of a post-deployment quality and safety program, including performance monitoring, human-in-the-loop review, and strategies to mitigate bias and over-reliance on AI.