The ups and downs of healthcare AI—so far

Have you heard AI is being used in healthcare? Duh, of course you have. But let’s dive into just how it’s being used.

The global AI healthcare market is expected to reach an $868 billion valuation by 2030, saving organizations a projected $646 billion and earning them $222, according to Strategy&, the international consulting firm of PwC.

But AI’s impact is more than just financial. There are also potential time-saving elements, drug discovery applications, and back-office efficiencies that, when utilized, can allow in-person care to flourish. There are also other areas of the industry that are ready for AI innovation, Aimee Cardwell, chief information officer and chief information security officer in residence at data privacy company Transcend, told us.

“We still want a human connection,” she said. “I don’t think you want to use Dr. Google or Dr. Claude for determining whether or not you have skin cancer. You really want to go to the dermatologist for that.”

Popular uses

Morning Brew Inc., recently conducted a survey of 277 healthcare professionals about healthcare AI usage, finding that the most common use of the technology at the moment is documentation or note-taking assistance. This includes tools by companies like Ambience, Abridge, and Suki that provide clinicians with ambient listening technology to create clinical notes in real-time during appointments.

Clinical documentation is considered one of the biggest burdens on providers and contributes to burnout. AI scribes saved 15,700 work hours, including “pajama time,” which is when providers finish up documentation at home, according to analysis from the journal NEJM Catalyst.

And according to these companies’ internal data, this aligns with what their AI tool users have claimed: Ambience reported 3.4 hours saved per day, Abridge reported users saved roughly two hours a day, and Suki users reported a 41% reduction in documentation time per note.

“There’s a lot of opportunity. The fact that there’s so much investment here shows how much opportunity there is,” Kavi Goel, director of product management for Health AI at Google Research, told us.

Other common uses of AI, according to Morning Brew’s survey, were for administrative tasks (45%) as well as voice agent and chatbot capabilities (30%). Some AI tools, for example, can answer phone calls to handle scheduling or streamline prior authorizations that get insurers to approve coverage of care—all of which work toward developers’ stated goals of saving time and reducing burnout.

In a 2024 survey of 1,000 physicians by the American Medical Association, 93% reported that prior authorization “delay[s] access to necessary care,” and 82% said it can lead patients to abandon treatment altogether.

AI tools can also help with clinical decision support like diagnostics and treatment recommendations, such as helping doctors read X-rays or connecting electronic health records data with patient data insights. Research or clinical trials were also areas where AI is commonly used, including using digital twins to simulate treatment outcomes. AI tools could save healthcare between $100 billion and $600 billion by 2050 through drug development, according to financial services company Morgan Stanley.

Downsides persist

The rise of AI usage in healthcare has brought concerns to the industry as well, however. For one, agentic AI poses cybersecurity risks in an industry already prone to hacking, with 48% of those surveyed by Morning Brew reporting it as their biggest concern.

There’s also a lack of federal regulation over healthcare AI, which has left industry groups to set standards, including the Coalition for Health AI (though Fierce Healthcare reported the information distributed by the organization has been confusing to some users as well as university resources.)

AI could also be used to deny patients care. Healthcare Brew previously reported on lawsuits against insurance companies allegedly using AI to deny coverage, though there are also AI tools that can now appeal such denials. One source told us at the time she was concerned that if there were more prior authorization denials then there could be more paperwork for hospitals.

On top of that are the environmental concerns. The United Nations reported a “global water bankruptcy” in January, meaning that many areas are living beyond their hydrological means. Considering large data centers that fuel AI can use up to 5 million gallons of water each day and healthcare is often the first line of defense against environmental disaster, the industry is unlikely to be immune from environmental challenges.

Looking ahead

In a January interview at the JPMorgan Healthcare Conference in San Francisco, Kimberly Powell, VP and general manager of healthcare with Nvidia, told Healthcare Brew the future of technology is “physical AI.”

“Imagine if any human on Earth could go get a lung screening scan just by walking into a room and being completely autonomously led through that experience,” she told us.

In recent weeks, major tech companies like Amazon, Anthropic, and OpenAI have released agentic tools patients can use to help them understand exam results on their own. These tools allow patients to ask medical questions directly to the bots and upload their patient records for analysis.

Healthcare Brew previously reported on a similar technology from OpenEvidence that is designed just for clinicians as well as a new integration between Microsoft Health and Wolters Kluwer Health to integrate its AI into clinician-facing tools.

While these tools are new, Goel said, for as long as Google has existed, people have been asking it medical questions.

“The potential value is extremely high, and the stakes are also high, so it deserves special focus to make sure we’re [using technology] responsibly,” he said.