AI-drafted patient portal messages do little for provider burnout

AI-drafted patient portal responses have the potential to help clinicians answer messages faster, but they don't fully address mounting provider workloads, according to new JAMA Network Open data.

The report examined the Epic In-Basket Augmented Response Technology at Columbia University Irving Medical Center and whether it supported faster patient portal message response times and less after-hours work for providers. These are key metrics as healthcare professionals work to manage growing patient portal inboxes.

Since 2020, patient portal message volumes have increased by 153%, according to an analysis of the Epic Cosmos database. That's resulted in an average of 2.6 hours of after-clinic work for healthcare providers, per a separate 2026 study in Health Affairs.

AI-drafted patient portal responses offer a potential solution to this problem. The AI can read through a patient's message and draft an appropriate response. Most healthcare organizations require clinicians to review the response before hitting send, but the idea is that AI would lessen the burden clinicians face in responding to an onslaught of patient queries.

But the situation is more complex than that, this latest JAMA report showed.

Notably, use of the AI response tool was linked with a 29.5% reduction in response turnaround time, meaning that when providers use the tool, patients can get their answers more quickly.

However, this data simply indicates that the tool works well when it's actually used. When looking at overall response times after the mere introduction of the tool, the researchers observed an 8.6% increase.

This is partly because an AI-drafted patient portal message comes with the obligation to review -- and potentially correct -- the response, which the researchers acknowledged adds to cognitive burden. In fact, the potential for added cognitive burden might have also discouraged AI use overall.

Additionally, the AI response tool did not yield any meaningful reduction in after-hours patient portal work. In fact, it was linked to an increase in after-hours work for messages received outside of clinic hours.

This could be because the AI lowers the barrier for providers to engage with their patient portal inboxes outside clinic hours. Prior to the AI, they might have held to a strict "stop time" for patient portal work outside the clinic. But with the AI, they might perceive their work as easier and become more flexible with that stop time. The result might be more patient portal responses, but it also means clinicians continue to let their work bleed into their personal time.

Still, the researchers said that their findings around after-hours work should be interpreted cautiously, as the data was not robust.

Although AI-drafted patient portal responses do not move the needle much on clinician burden, the researchers said it's not without merit. The technology can speed up response times to low-level messages that require straightforward provider review. Although patients won't get quick responses to every message, the AI could boost satisfaction with more administrative queries, like prescription refills or appointment booking.

Still, health systems will need to look further to address clinician burden.

Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT.