Some physicians save an hour a day with AI scribes — 42% save no time at all

Physicians remain split on whether artificial intelligence (AI) is improving medicine, even as adoption of AI scribes climbs across primary care and specialty practices. Among physicians already using the tools, real-world results are uneven, according to results of a flash poll presented April 29, 2026, during a Medical Economics and Physicians Practice webinar sponsored by Heidi Health.

The poll, which gathered responses from 103 physicians, showed that 55% believe AI is improving medicine, whereas 45% do not. Seventy percent of respondents currently use AI scribes or are evaluating vendors, but the time those physicians report actually saving varies widely.

Forty-two percent said the technology has saved them zero minutes on a typical clinic day, whereas 20% reported getting more than an hour back.

The webinar was hosted by Chris Mazzolini, vice president of content at MJH Life Sciences, parent company of Medical Economics and Physicians Practice. He presented the data alongside Nancy Cibotti, M.D., a primary care physician and U.S. chief medical information officer at Heidi Health.

Why time savings with AI vary so widely

The 42% reporting zero minutes saved was notable, given that documentation relief has driven much of the early enthusiasm for ambient AI. When asked what excites them most about AI tools in medicine, 32% of respondents pointed to reducing documentation time through tools such as AI scribes — the most common answer.

Cibotti, who previously oversaw an AI scribe rollout across Beth Israel Lahey Health before joining Heidi earlier this year, said the variance in real-world time savings does not surprise her.

She pointed to a recent multicenter study led by Lisa Rotenstein, M.D., MBA, an associate professor, physician and researcher with the University of California, San Francisco (UCSF) Department of Medicine, which examined ambient AI scribe use across five academic systems: Mass General Brigham, UCSF, UC Davis, Yale New Haven Health and Emory Healthcare.

“Overall, they saved about 16 minutes of EHR documentation time, 13 minutes in total, which isn’t great,” Cibotti said. In that study, each of the systems used Ambience, DAX Copilot or Abridge, and Epic as their electronic health record (EHR) vendor.

Her own experience tells a different story. At Beth Israel Lahey Health, the rollout of Heidi’s AI scribe tool saved an average of 3.8 minutes per visit, with longer or more complex specialty visits saving up to 6 minutes, she said. “It makes me wonder, especially for the [Rotenstein-led study], are they using the right tool, and are they using the right tool in the right way?”

Cibotti said it’s a misconception that all AI scribes work the same way. “They’re really not [the same],” she said. “If you are spending a lot of time either correcting misinformation or reorganizing a note, you’re not saving any time at all.”

Cibotti also said many physicians use the tools too narrowly, treating them only as in-visit transcription rather than a broader documentation engine that can extend to prior authorizations, patient letters and clinical decision support workflows.

The patient interaction question

A separate question in the survey asked whether AI scribes had affected the quality of physician-patient interactions during visits. Forty-one percent reported no noticeable change. Twenty-four percent said the interaction had improved significantly, with more eye contact and less typing, and 20% said it had improved slightly. Fifteen percent of respondents said the experience had worsened with the implementation of AI scribes, due to distractions or inaccuracies.

Cibotti said the upside hinges on physician trust in the tool’s accuracy.

“It really comes down to trust and accuracy,” she said. “If I don’t trust the tool, I’m still going to be taking notes on the computer. I need to be able to know that it’s accurate so that I can put my computer aside and really listen and talk to the patient.”

She added that one underdiscussed benefit is the way scribes preserve the patient’s own words inside the note.

“Often, with all the heavy templating that we do, we lose the patient voice,” Cibotti said. “Now the scribe actually captures that patient story in a way that I think they appreciate it.”

The financial case for AI scribes

Bottom line is that 53% of physicians said it’s too early to determine the financial impact of AI scribe adoption. Another 17% reported costs were outweighing measurable financial benefit, 16% said they had come out even and 14% reported a clear positive return on investment (ROI).

Cibotti cautioned against framing the financial argument primarily around fitting more patients into the day.

“When scribes first came out, people were very excited over the idea of, ‘Now we can see more patients because we have time saved,’” she said. “But I think that is a very limited ROI, and I would caution people about approaching these tools in this way.”

She pointed instead to physician retention and reduced burnout as the more durable financial argument. At Beth Israel Lahey Health, Cibotti said, the simultaneous rollout of Heidi alongside Epic implementation kept some clinicians from retiring early.

“I had clinicians say to me, ‘You know what, I don’t want to deal with Epic, I am just going to retire early,’” she said. “And I would say, ‘Wait, let’s try this tool in addition.’ None of those clinicians retired.”

Looking past the scribe

When asked what physicians should look for in a vendor, Cibotti said flexibility across specialty workflows matters more than tight EHR integration. Her own thinking on integration, she said, has changed.

“I used to think integration was the top priority. I no longer think that,” she said, noting that Heidi rolled out at Beth Israel Lahey Health as an unintegrated product before later connecting with Epic and Athena. “It really is the clinician experience with the tool and the ability to be flexible.”

She urged practices to think beyond the scribe altogether.

“You really shouldn’t be looking for an AI scribe. You really should be looking for an AI solution,” Cibotti said. “We’ve gone well beyond scribes now, whether it’s clinical decision support or agentic agents.”

The survey’s data on physician concerns suggests practices are still working through that broader question. Asked what makes them most nervous about AI, 32% pointed to AI replacing clinical judgment or disrupting the physician-patient relationship, 23% cited liability if AI produces incorrect recommendations and 20% flagged patient safety risks from inaccurate or biased outputs.

Thirty percent of respondents said their practices are not using AI scribes, nor are they considering them — a non-adoption rate Mazzolini noted during the webinar is similar to early physician resistance to EHRs during the Meaningful Use era.

Cibotti said the best AI tools are designed to keep the clinician in the loop, particularly because patient context often falls outside of what the technology can capture.

“There may be some cultural beliefs or [social determinants of health] factors that we need to take into consideration,” she said. “There may be other things that are going on in that patient’s life that … are not necessarily verbalized during the visit.”

Looking ahead, Cibotti said she expects scribe technology to fade into a broader, more invisible layer of clinical AI handling between-visit follow-up, medication adherence checks and other tasks primary care has historically struggled to do at scale.

“I think we’re going to stop talking about scribes,” she said. “We’re just going to talk about AI.”

Related content: Take note: The AI scribe era is here