Patients prefer healthcare providers' AI agents to public chatbots, with human oversight non‑negotiable, survey finds

Hand using smartphone and chatting with chatbot application, shows message speech bubbles digital screen of social media noti
Forty-four percent of patients say a 24/7 agentic assistant would make them more likely to stay within a provider’s network for follow-up care, a Salesforce survey found. (patcharin innara/GettyImages)

Patients are open to using AI agents to help with tasks like scheduling and billing, but they overwhelmingly prefer AI tools that live inside their doctor’s secure portal, a new Salesforce survey found.

Patients are three times more likely to trust an AI agent when it’s embedded in a clinical system rather than offered as a public chatbot. It's a clear signal that institutional accountability and provider context are central to acceptance of healthcare AI, according to Salesforce's survey.

The company surveyed more than 3,200 patients across eight countries to understand how the rise of agentic AI is reshaping consumer expectations. Salesforce surveyed patients across the United States, Canada, Mexico, Brazil, the United Kingdom, Ireland, Australia and New Zealand from March 24–April 10.

In early 2024, just 2% of U.S. adults turned to AI for healthcare information, according to a Salesforce survey from two years ago.

Today, 61% of global patients say they are comfortable using agentic AI in healthcare contexts, and 64% would share their full medical history with AI for a faster diagnosis. And that mirrors what's happening on the provider side: a separate Salesforce study found 71% of U.S. healthcare workers predict agentic AI will be essential to healthcare operations within five years. 

Patients face ongoing friction and pain points when it comes to accessing care. Forty-six percent of patients delay care because the digital process is too confusing, and 58% skip necessary care because scheduling is too difficult. Nearly half (49%) hang up after 10 minutes on hold with a doctor’s office to seek care elsewhere or avoid it altogether. Two in three (66%) have run out of medication waiting for a prescription refill to be approved. And 90% say they wish their primary doctor was automatically notified after an ER visit, according to the survey.

“We all know that patients are utilizing AI in a variety of ways in everyday life, but especially when it comes to healthcare, because I think healthcare is really difficult for patients to navigate and to understand what is the right provider, what are my next steps, what should I do with the symptoms that I might be having, and so it's understandable that patients are reaching out to whatever tools that they have to be able to really understand what their next steps are,” Tyler Bauer, senior vice president of system ambulatory operations at UChicago Medicine, told Fierce Healthcare.

Advances with agentic AI open up opportunities for providers, according to Salesforce's report, as AI tools can extend support beyond traditional business hours, reduce administrative friction and improve access.

But patients view governance and provider trust as non-negotiable prerequisites, and they don't want AI acting alone.

Nearly 9 in 10 patients (89%) say a clear “escalate to human” option is essential for trusting AI administrative support, and 90% expect the same for AI medical support. And 91% say patients should have the right to opt out of AI-driven clinical recommendations entirely.

Patients’ top concerns about AI in healthcare are accuracy, followed by data privacy. Thirty-six percent of patients cite accuracy of diagnosis or treatment as their primary worry, while 30% point to the privacy and security of their health data.

Privacy, security and compliance are foundational priorities for healthcare organizations. "Trust and transparency are at the foundation of patient experience. We're not going to compromise on that," UChicago Medicine's Bauer said.

"Think about your relationship with your physician that is really based on trust and a relationship which I think healthcare does a really terrific job of. Patients have had the opportunity to work through patient portals for a number of years, so that is familiar to them as a secure and compliant way to access medical information. I think that trust that we have worked so hard to build with patients will continue even in this world of additional new technology that is coming out through AI," he noted.

The data shows that patients are increasingly comfortable with AI handling logistical and navigational tasks like billing and rescheduling, especially after hours. Nearly half of patients (49%) prefer AI agents over humans to avoid wait times, but only when human backup is visible and accessible. Sixty-seven percent say they would rather have 24/7 AI help than wait for someone to pick up the phone during office hours. 

“Patients don’t want AI to replace their doctors. They want it to safely replace the waiting, and the friction. When technology is built on trust, healthcare can finally move as fast as we do," said Sophia Saleem, M.D., Chief Health Officer, HLS Go To Market, at Salesforce.

"Health systems need to be ready to utilize tools that are out there but do it in a way that fosters the trust that patients have with the health system, and they feel that their information and their data is secure and their conversations that they're having are secure as well. That's exactly what we are working on and implementing at UChicago Medicine," Bauer said.

He added, "We're on a mission to make access and care easier, seamless and friction-free for patients and so our focus with AI is on administrative and access-related interactions, certainly not replacing the clinical judgment of our physicians or our care teams."

UChicago Medicine's access center, its customer service hub, fields close to 3 million inquiries a year. "We took a look at the data, and we found initially seven use cases that are calls that are maybe repetitive information that is easily out there that our team members are having to repeat over and over again. We spend a lot of time training our team members on how to schedule and work through complex issues with our patients, so we found an opportunity to automate some routine inquiries so our team members can spend more time helping patients with complex needs and higher-touch interactions," Bauer said. "We've initially focused on those seven use cases and have several of them live so far."

Working with Salesforce, UChicago Medicine is using AI agents to answer webchat inquiries about directions to a clinic or facility and appointment confirmations.

"We're about to go live with canceling and rescheduling primary care appointments as well for our patients, but other plans include some additional scheduling for patients, including new patients, and capturing and routing medical advice requests to the appropriate clinical team so that they can respond and then helping patients navigate through prescription refills," Bauer said.

The use of AI agents can provide consumer-grade service for patients, on par with what consumers expect from other digital experiences.

"Salesforce has provided us a platform that helps us connect to our patients through patient engagement channels, manage those workflows and support consistent experiences across digital and human touch points, but also allows us to implement automation within a framework that incorporates the privacy, security and operational requirements that we have in place," Bauer said.

The use of AI tools to reduce friction can boost patient loyalty, the Salesforce survey found.

Forty-four percent of patients say a 24/7 agentic assistant would make them more likely to stay within a provider’s network for follow-up care, and 65% of patients with long-term conditions say a 24/7 digital helper would make managing their health significantly easier. 

The survey findings highlight that patients are ready to embrace agentic AI, but only when it is built on transparent, governed foundations with clear escalation paths, audit trails and provider-backed deployment, according to Salesforce executives.

"Technology is advancing quickly and innovation seems to be moving faster, potentially faster than some of the regulatory frameworks, but our view is that healthcare organizations have a responsibility to engage with these technologies thoughtfully, and in alignment with evolving regulatory requirements, rather than wait on the sidelines," Bauer said.

There are also opportunities to use AI agents for more proactive outreach to patients, Bauer mentioned, to close care gaps for things like reminding patients to get an A1C test or to get a vaccine. "Being able to automate that through this technology I think is a huge win for our patients, as well as our physicians and our team members who support them," he said.