The doctor’s AI will see you now

Doctors are drowning in paperwork. Studies show they spend over a third of their workday (plus at least two hours per day outside of work) on clinical documentation, contributing to widespread burnout. But maybe it doesn’t have to be that way; a burgeoning industry developing AI medical documentation tools — or “ambient scribes” — and a growing number of physicians think this busy work should be delegated to AI.

The tools were originally designed to draft medical notes. Now companies are increasingly pushing them further. This includes providing context on the patient’s medical history, assigning billing codes and even acting on the next steps of a patient’s care.

“We can help not only with the notes, but we can hear your orders, so we can place your prescriptions or your diagnostic orders for X-rays or CTs. We can set the table for you on all these other tasks,” said Shiv Rao, founder and CEO of Abridge, a startup that in 2025 doubled its valuation to $5.3 billion within just four months.

The category has grown quickly, from a few early-stage products to a market where startups, tech giants and healthcare companies all compete. Investors poured at least $1 billion into the space in 2025 alone, and Menlo Ventures in its 2025 State of AI in Healthcare report called ambient medical scribes “healthcare AI’s first breakout category.” (Menlo is invested in medical scribe company Heidi).

It has all the makings of a major boom, but the path to adoption isn’t anywhere near as simple as spinning up and funding products. The AMA’s 2026 Physician AI Survey found the adoption rate of the tools lags behind clinician enthusiasm for trying them — perhaps not surprisingly, placing AI recording software in between patients and their doctors comes with a lot of hurdles. While the tech and healthcare companies selling AI medical scribes have slightly different paths to market share available to them, they’re both up against rigid procurement processes and a whole lot of trust issues.   

Trust and process

According to Wiljeana Glover, a researcher focused on healthcare innovation at Babson College, the gap between adoption and interest is mainly down to the procurement process itself. Bringing on such a technology is a big decision in these settings, and medical groups have very formalized processes to identify their needs, select potential candidates, screen the top choices and evaluate them on specific criteria. “That process can take time,” she said. 

Early studies suggest the tools are effective at reducing clinician time spent on documentation, but they’ve been limited to one healthcare system or have been conducted over short periods of time. Jamie Chang, chief medical officer at WellSky, a healthtech company with an AI medical scribe product, said the challenge is getting the decision makers to trust the tool will be accurate and won’t create any additional burdens.

“I think the value prop and desire is there, but maybe [it takes] more trust in the process,” he said. 

To sell to big medical groups, Rao described having to go through several executives including the chief medical officer (CMO), CFO and CIO. The CMO wants to know specifically what problems you can solve for their clinicians, while the CFO is interested in the bottom line, he said. The CIO needs to be assured you meet all the privacy and security requirements, and will be able to serve them for years to come. At the same time, selling to providers outside those big health systems comes with its own challenges, in particular just trying to get in front of them. 

“The go-to-market for those clinics can be very painful,” he said. “It's hard to knock on all those doors and be able to scale something quickly. So finding channels for distribution becomes really key, especially for startups who are fighting incumbents or hegemons.” 

With companies of starkly different sizes from both the healthcare and tech industries offering these products, Glover said she’s seen their ability to effectively go through the procurement process vary. She’s also noticed a pattern of the large health systems choosing for consistency, even if it means sacrificing on customizability of the tool.

WellSky, for example, is now giving away its AI medical scribe offering for free to customers already using its platform for other functions, which Chang said is “absolutely” an advantage as options flood the market.

“I think over time, the industry continues to consolidate rather than have a lot of bolted-on options,” he said. 

Patient perspective

Ambient medical scribes have potential benefits for patients, too: a doctor who isn’t heads-down taking notes during your appointment or generally burned out can theoretically provide better care. Yet this potential upside isn’t obvious to patients and doesn’t necessarily outweigh concerns about AI recording their sensitive medical conversations, especially as overall distrust of AI increases.

In one survey of patient attitudes toward ambient medical scribes conducted at Stanford Health Care in 2025, 73.6% of 2,202 patients who experienced their clinician using it said they would like it to be used again. They were told in advance that the visit would be recorded and given the chance to decline. Without first-hand experience and upfront disclosure, however, patient willingness looks different. A much larger survey of 12,153 Canadian adults based only on perception and not a direct experience found that only 28.3% were aware clinicians were using AI scribes at all, and that, even while anticipating benefits from its use, 61.8% were reluctant to have an AI scribe used in the future. The researchers cited privacy concerns and low awareness as key barriers.

“The ethical risks fall disproportionately on patients... [Ambient AI systems] generate vast amounts of data that may have secondary uses — whether anticipated or not — raising significant concerns about privacy, surveillance and data governance,” wrote the authors of a 2026 comment on consent and ambient clinical AI.

The authors also argue current deployment practices fall short of getting meaningful patient consent, including not adequately explaining recording practices or providing opportunities to opt out. Several healthcare providers have already been sued by patients in class action lawsuits claiming the technology was used without consent.

On patient privacy, Rao said a lot of Americans now “just expect that things are getting captured in some way.” and that AI meeting note-takers have normalized the idea. Doctors still need to get consent and spend time talking to patients about it, he said. Chang noted physicians he’s worked with are facing patient questions about how the recordings are used. This is putting doctors on the spot to talk about AI and data practices they might not even fully understand themselves, said Glover, adding that the tools should be designed with privacy features top-of-mind, such as automatically filtering out sensitive data.  

“Keeping patient needs front and center allows health systems to truly make wiser decisions,” she said. “The influx of a lot of shiny objects, as we might call them, can become really distracting.”