AI Can Diagnose, Prescribe, and Decide. Is it Time to Replace Clinicians?

The ‘Her’ Paradox

For the consumer/patient, chatbots are becoming an effective tool to gather information, schedule medical appointments and get test results. Some providers are experimenting with chatbots to help with primary care coordination, identifying through questions what care a patient needs and routing them to the right provider.

But here’s where an interesting conflict surfaces. Studies have found that while consumers have a fear of talking to an AI bot instead of a real person when seeking healthcare services, some have also found that an AI bot can be more sympathetic than a live person.

“Although AI doesn’t experience emotions, AI-powered tools, like chatbots, can offer empathetic-like responses, at all hours of the day,” Elizabeth Grill, Psy.D., an associate professor of Psychology at Weill Medical College at Cornell University, wrote on the Psychology Today website. “Many patients appreciate the chance to ask questions or get medication reminders without waiting for office hours. In fact, some studies even show that, in certain cases, patients perceive these AI-driven interactions as more empathetic than conversations with rushed human providers.”

That can be especially helpful when screening for sensitive topics like substance abuse, sexual diseases and violence. Patients might feel more comfortable disclosing the truth to an AI bot, such as how many drinks they’d had, whether they’re used drugs or whether they’ve had sex.

Patients are sometimes “willing to tell a more complete story of their health,” says Crowder, at Advocate Health. “They don’t feel judged.”

The question then is whether bots can be ‘too human,’ or whether patients can rely too much on what AI gives them – much like Joaquin Phoenix’s character in the 2013 film Her, who falls in love with an AI bot (and if you think that’s just a movie, a recent survey by Vantage Point Counseling Services found that more than one in every four adults say they’re had an intimate or romantic relationship with AI, and more than half claim to have had some kind of relationship with an AI system, whether as a friend, colleague or confidant). Some states, like Illinois, are strictly regulating or even banning AI bots in mental healthcare because of concerns that they could harm patients who rely on them too much.

Blum notes that AI bots can also help providers cut back on the 15-20 phone calls they’d be making every day just to follow up with patients. 

“Humans are human, and being trained to speak to the least common denominator isn't necessarily something that we're capable of,” he points out. “Nor do we want that.”

What executives do want, he says, is an AI tool that can communicate with patients in a clear and pleasant manner, exchanging data with them and giving clinicians what they need to manage and improve care. When that care is delivered, however, is when the clinician steps in.

“Our job is to do what's best for patients,” he says. 

Crowder agrees. Agentic AI tools can do the heavy lifting on administrative tasks and data crunching, can conduct initial interviews with patients and even direct them to the right providers, and they can connect the dots and fill in the gaps. But they can’t deliver healthcare.

Chaudry, of Seattle Children’s, says he’s not convinced that AI bots can or should replace humans when communicating with patients. Empathy may be one thing, but talking to a real person—especially in healthcare—matters more.

“I can get a clinician to a [patient] within 70 seconds or less, and they can talk to a real person [and] get their problem solved 90% of the time without using AI,” he says. “And when I cashed it out, it’s cheaper than building an AI tool.” 

Eric Wicklund is the associate content manager and senior editor for Innovation at HealthLeaders