How can generative AI meaningfully support clinicians while improving communication in women’s health? A new study from UC San Diego Health, led by Ghanshyam Yadav, MD, published in O&G Open, begins to answer that question.

The study, ​“Evaluating Generative Artificial Intelligence – Drafted Responses for Patient Messages in Obstetrics and Gynecology” is one of the first investigations to examine how GenAI can be integrated directly into an EHR’s patient messaging workflow within an academic OB-GYN department. The study, in partnership with the Jacobs Center for Health Innovation, offers valuable, real-world insight into where AI can help, and where thoughtful oversight remains essential.

For Dr. Yadav, the inspiration was simple: clinicians spend too much time behind computers and not enough time connecting with patients. Patient portal messages, while essential, often become an overwhelming after-hours workload.
The question was whether generative AI could help shift that burden, turning ​“more typing” into ​“more talking.”

“As physicians, we entered medicine to connect with and help patients, but increasingly we find ourselves spending more and more time behind a screen,” he said. ​“There was a clear gap between the excitement around AI tools and the lack of real-world data on how they could be integrated responsibly into clinical workflows. Our study was designed to bridge that gap. Ultimately, this work came from a passion to better operationalize AI in healthcare and to turn it from a buzzword into something that tangibly improves how we deliver care and how clinicians experience their work.” 

The team integrated GPT‑4 to generate draft responses to patient messages directly within the EHR. Clinicians remained fully in control, reviewing, editing, and personalizing each draft, but the AI served as a smart, embedded assistant.

Key findings of the study include:

  • Feasibility: GenAI-assisted drafts were successfully integrated into the EHR workflow.

  • Reduced Workload: Clinicians reported lower perceived workload and modest improvements in burnout.

  • Consistency & Compassion: AI helped maintain a warm, empathetic tone, something clinicians naturally convey, but can struggle to express under time constraints.

  • Human Oversight Is Essential: Participants emphasized the need for specialty-specific tuning, transparency, and careful workflow design.

“I think our findings underscore that AI in healthcare isn’t just about efficiency, it’s about also attempting to restore the clinician by reducing digital workload so that they regain time and emotional bandwidth for what AI cannot help with,” Dr. Yadav said. 

Looking ahead, Dr. Yadav emphasizes that women’s health, historically underfunded and underrepresented in research, stands to benefit enormously from responsible innovation. 

“Much of the data that drives medicine today was never built with women in mind,” he said. ​“That’s the gap I hope technology and AI can begin to bridge. But this progress won’t happen automatically. We have to be intentional in designing systems that are effective, transparent and guided by empathy. If we do that, innovation can start to correct decades of imbalance, helping women be better represented, better understood and better cared for.”

This study marks an early but important step toward understanding how generative AI can responsibly support clinicians in women’s health, improving the experience of care for both patients and providers.

Study co authors: Kshitij Pandit, MD; Erin A. Gross, MD; Karandeep Singh, MD, MMSc; Yvette D. LaCoursiere, MD; Ming Tai-Seale, PhD, MPH; Marlene Millen, MD; Cynthia Gyamfi-Bannerman, MD, MS; Sarah Parker, BS; Divya Gupta, MS; Christopher A Longhurst, MD.