As artificial intelligence gains momentum in healthcare, a seemingly contradictory question is sounding alarms in the minds of skeptics and supporters alike: Could AI tools designed to ease clinicians' burdens actually increase their workload?

Given that the technology is still in its early stages, it's too soon to draw conclusions. But experts say the answer—which carries significant implications for AI's future adoption in the industry—depends in part on how often clinicians need to verify model outputs, as well as the financial incentives within healthcare's predominantly fee-for-service payment structure.

'The tools are not perfect'

Many healthcare AI tools aim to reduce administrative friction by helping with routine tasks such as documenting patient visits, finding medical information, or filling out paperwork. Such tasks contribute to clinician burnout, a major issue in the industry that can drive physicians out of the field.

In a 2023 American Medical Association survey, about one-third of responding physicians expressed interest in or planned to leave the profession within a few years. Many cited overwork as a reason: according to some studies, healthcare workers spend most of their week on administrative tasks.

Enter AI. Algorithms have been used in healthcare companies for decades, but in the past few years, excitement around the technology has reached a fever pitch with advances such as generative AI, which can create original text, and AI agents, which can perform tasks without direct supervision.

Many companies are integrating the technology into polished products, touting time-saving metrics that may seem too good to be true for clinicians tired of doing documentation during 'pajama time' after hours.

Abridge, a startup that uses ambient listening and generative AI to automate clinical notes, and Nabla, another ambient AI assistant, both claim to save physicians about two hours per day.

Software giant Oracle has integrated its AI into its health record platform for physicians, including a clinical AI assistant. According to the company, doctors using the tool spend 30% less time on documentation.

Meanwhile, Microsoft says its AI documentation product saves physicians five minutes per patient visit on average.

'The capabilities that AI scribes bring to medical practices are remarkable,' said Rohit Chandra, chief digital officer at the Cleveland Clinic, during a panel discussion. (The academic medical system is currently rolling out Ambience's AI documentation software across its physician network.)

Documentation products address a pressing pain point for physicians while being relatively easy to use, simple to implement, and safe because they are overseen by clinicians, Chandra said.

However, experts say AI performance metrics, such as time saved, should be taken with a grain of salt.

Clinicians still need to review AI outputs for errors. This process, known as 'human-in-the-loop,' is key to ensuring accuracy and building trust in AI, experts say.

Letting AI tools run without supervision could lead to fabricated symptoms in medical records or missing information critical to a patient's health.

'I think that's where we could get into trouble. The technology will get to a point where, at least in my view, we won't have to worry as much. But in the early adoption phase, we have to be aware that it will make mistakes,' said Brenton Hill, head of operations at the Coalition for Health AI, an alliance for health AI standards, during a panel discussion.

But having clinicians oversee AI-generated content could significantly cut into time savings, in some cases eliminating them altogether.

'The tools are not perfect,' said Deborah Edberg, a family physician at Oak Street Health, a primary care chain owned by CVS, during a panel discussion. 'We use AI for documentation, and I do spend a fair amount of time going back to edit... Ensuring the notes are accurate can become a burden.'

To date, healthcare AI scribes have not undergone extensive independent review. But a recent study published in JAMA Network Open found that Microsoft's ambient scribe received mixed feedback from users.

'A recurring theme was the need for substantial editing and proofreading of AI-generated notes, which sometimes offset time savings,' the researchers wrote.

Time savings may not benefit clinicians

Fee-for-service incentives could also be a problem. This payment structure underpins most U.S. healthcare spending, rewarding physicians based on the number of visits, tests, and procedures.

Meanwhile, hospital operators' costs continue to rise, even though operating margins have recovered somewhat since the COVID-19 pandemic.

As a result, hospitals and other physician owners have an incentive to schedule clinicians as tightly as possible to bring in as much revenue as they can. Any time AI frees up for physicians could be filled with another visit, experts say.

'That could happen. Some organizations want to do that. From our perspective, it's an organizational decision,' said Rachel Wilkes, enterprise lead for generative AI programs at EHR vendor Meditech. 'If they want to use the time saved to schedule more patients, they can do that.'

Measures to increase productivity in health systems—such as reducing patient no-shows—often lead to heavier physician workloads, said Seth Howard, vice president of R&D at Epic.

'The conversation between health systems and physicians needs to be: 'How do we handle this time saved?' Is it giving physicians some flexibility during the day... or is it having physicians see more patients?' Howard said.

Vendors are aware that AI products' appeal to hospitals includes the potential to increase revenue.

In addition to saving five minutes per visit, Microsoft's ambient AI 'frees up 13 additional appointment slots per physician per month,' according to marketing materials.

The need to shore up weak finances is one reason hospitals are eager to use these tools, experts say. Demand has been soaring, with one CEO of a company using AI to automate physician workflows comparing it to a 'fairy tale.'

'Our ambient product went from zero bookings two years ago to hundreds of thousands last year, and this year we're on track for 10 million,' said Tanay Tandon, CEO of Commure, during a panel discussion. 'The expansion and use of this AI in healthcare is incredibly exciting. This is no longer nerdy tech doctors tinkering in a corner. This is meaningful use.'

Experts say that, like any other technology, AI scribes, assistants, and similar tools have pros and cons. Moreover, beyond time savings, other metrics matter to physicians, such as reducing mental burden by streamlining tasks or improving job satisfaction by allowing them to focus on clinical care.

Anecdotal feedback from clinicians has been very positive, according to tech giants, AI scribes, and EHR vendors that have integrated the technology into their workflows.

For example, Epic doesn't have an average time-saving figure for its ambient documentation tool. But according to Howard, 'It's very common for physicians to say, I could never go back to the way I documented before.'