By the numbers: What’s the current state of AI in healthcare?

It’s going too fast. It’s going too slow. Clinicians use it for everything 24/7, but most pilots never make it into their hands. Patients love it, except when they hate it.  It’s definitely saving everyone money, but no one has the budget to make it happen at scale. 

What the heck is actually going on with AI in healthcare?  Well, it’s definitely complicated – and the real-world enthusiasm and success might not be quite as universal as what your LinkedIn feed is signaling. 

In the past few months, a flurry of surveys have been released that, taken together, paint a more nuanced portrait of how, when, and where AI is finding its place in healthcare – and how far the industry still has to go to find true value in some of its promises. 

Adoption is definitely increasing, spurring notable business changes 

One thing we can say for certain is that the AI arms race is well and truly underway. Adoption rates have spiked in the past several years, with the latest data from NVIDIA pinning active healthcare sector adoption at 70%, up from 64% in 2024. 

A HIMSS and Guidehouse survey from the summer of 2025 shows even higher numbers, with 78% of respondents saying they have active AI projects underway with established plans to keep expanding their AI efforts. More than half (58%) are looking to increase automation via AI in the next two years, if they can overcome lingering barriers around strategy, security, and data accessibility.  

Health systems – and their employees – are already seeing some shifts due to the AI infusion.  The NVIDIA survey says that 85% of executives believe that AI is helping to increase revenue, and 80% reported that the technology is reducing costs.  

Data from Morgan Stanley shows that AI is also helping to increase productivity, with healthcare leading other industries with a 12% net productivity increase. 

But these gains are already coming with a human cost, Morgan Stanley added. Globally, the healthcare sector has experienced an intentional 5% net loss in employees due to a higher number of eliminated roles than new hires. Entry-level and early career workers are most often affected, while more experienced employees are more often retrained or reassigned. 

Physicians are getting more comfortable with using AI – with some reservations  

Physicians have always had mixed opinions on the role of new technologies in their practice, and AI has been no exception. Overall, trust has been slow to materialize, especially as emerging real-world stress tests and vendor AI-washing continue to breed skepticism over claims about safety, quality, and value. 

But more and more physicians are embracing the inevitable and getting used to AI in practice. More than half of physicians (54%) told athenahealth that they’re “comfortable” using AI in their daily workflows.   

And shockingly, AI might finally be turning the tide on widespread dislike of using the EHR, with 62% of physicians now agreeing that the EHR makes them more efficient (up from 54% in 2024) and 59% saying that the EHR actually simplifies clinical workflows. 

However, doctors remain deeply concerned about the overall impact of AI on the profession, with 6 in 10 saying they are worried about AI contributing to a loss of the human touch in healthcare.  This number is unchanged from 2024 and 2025. Close to a third (30%) also said that they feel “AI will be one more thing that complicates healthcare,” although that’s down from 46% in 2024.  

Physicians practicing in less digitally mature organizations, especially those still facing significantly interoperability, data access, and care coordination challenges, were less likely to feel like AI is a useful tool, the athenahealth survey found. Clinicians in these environments were dramatically more likely to express discomfort with using AI for diagnostic support and treatment planning, and showed less trust in using AI tools for administrative tasks. 

This indicates that health systems seeking to increase trust and real-world adoption will need to start by improving the fundamentals of data accessibility and information exchange.  

Skeptical patients are demanding transparency and accountability 

If physician opinions have been mixed, patient opinions have pretty much always been clear.  Trust is even less common among consumers, and calls for transparency and accountability have been loud and steady since day one. 

The desire for clear and trustworthy AI use in healthcare is only getting stronger as users get to grips with what it means to have automated tools help with decisions about their care, reveals a January report by CHAI, NORC, and the California Healthcare Foundation. 

Only 13% of consumers say they’re “comfortable” with AI in general, and a whopping 93% of consumers expressed at least one concern with the use of AI in healthcare. More than four times as many respondents say AI use makes them trust healthcare less than before (51%) rather than more (12%). 

More than 80% of respondents said they’re worried about lack of oversight, and a similar number said there are risks that clinicians could rely too heavily on AI instead of making their own judgments.  

Flipping the script won’t be easy, but there are a few things health systems can do to boost trust: prioritize transparency around AI use, ensure that humans remain in the loop, adopt strong oversight and validation processes, and enable patients to opt in instead of opt-out of AI-driven care.  

For example, more than 80% of respondents said they want to know when AI is being used to assist with decision-making or treatment planning, including the use of AI for coverage determinations and prior authorizations. About 60% say they want to be explicitly asked for permission for AI to be used for these functions and given the opportunity to say no. 

Some of these concerns stem from the recognition that personal health data is a commodity to AI companies. There is widespread awareness that personal data is being used to train algorithms, and equal levels of discomfort about it. Close to two-thirds (63%) are somewhat or very concerned that their data is being sold or shared to make money, and 69% believe that patients should retain ownership and control of their data in the AI ecosystem.  

It’s going to become increasingly complicated to implement data controls and opt-in processes as AI continues to make its way into every corner of the healthcare experience.  But leaders can still focus on other key demands from consumers, such as meaningful oversight from trusted third parties and responsible use of disclosures, to help diffuse distrust and encourage engagement.  

The big picture 

AI adoption rates are well past the tipping point, but that doesn’t mean clinicians and patients are entirely on board. Health systems leaders still have a lot of work to do around demonstrating the types of real-world value that can increase trust among both clinicians and consumers. And they’ll need a multi-pronged attack to do so.  

To convert physicians, leaders will have to keep hammering away at the data siloes that restrict full visibility into patient care so that AI tools have the right data to work with.  And to woo consumers, they’ll have to double down on governance to ensure that their AI tools are functioning safely, transparently, and with the patient’s best interests in mind at all times.  

Both of these are critically important priorities that will work together to strengthen the entire AI adoption process, hopefully leading to continued gains in productivity and cost savings with minimal risks to experiences and outcomes. 

Jennifer Bresnick is a journalist and freelance content creator with a decade of experience in the health IT industry.  Her work has focused on leveraging innovative technology tools to create value, improve health equity, and achieve the promises of the learning health system.  She can be reached at [email protected].