3 considerations for health systems adopting AI chatbots

1. Health systems must be proactive in AI adoption

According to early chatbot adopters, the biggest mistake a health system can make is to not consider patient-facing AI at all. The commercial tools are here to say, so it's a matter of whether or not organizations keep a hold on their patient relationships, according to Ran Shaul, co-founder and chief product officer of K Health.

K Health is a clinical AI company that's worked with a handful of health systems to launch their own AI chatbots, such as Hartford Healthcare's PatientGPT.

"Health systems are not going to stay passive," Shaul said.

They've made that mistake before when they were too slow to react to the GLP-1 boom, he added. Now, digital health companies have most of the power in that space and health systems are left to contract with those vendors.

Health systems won't repeat that dynamic, Shaul explained. By launching their own chatbot tools, health systems are maintaining ownership of patient relationships, which will be the ultimate KPI for these technologies.

"The PatientGPT success will be that we become an everyday life aspect of our patients," as opposed to the reactive, sick care the industry has been known to provide," Shaul said. "We want to become part of your healthcare and be connected to you. We want you to ask anything you want, and to think about preventive care and the next steps you need to take."

For Muhammad Siddiqui, M.D., CIO of the Indiana-based Reid Health, that message is even more salient when there are off-the-shelf products available. His health system adopted Ask Emmie, the AI chatbot product from Epic Systems. It's allowed his rural health system to stay on an even playing ground with bigger academic medical centers.

"Some organizations are too slow to respond to these features that are coming out from Epic," Siddiqui said.

Since implementing Ask Emmie, the technology has reduced the number of administrative tasks for humans and helped streamline the patient experience.

"That's exactly our goal: how can we make life easier for our patients?" Siddiqui posited. "Patients are looking for speed and agility to simplify things, as well. Healthcare systems really need to do their due diligence, lay out the plan and have a solid roadmap to roll out these features in AI."

2. Be mindful of existing provider workflows

Health systems can't be passive as AI chatbots become more popular, but they also can't adopt AI for AI's sake. Rather, healthcare organizations must determine how they want AI to supplement existing workflows, according to Chetan Aher, M.D., associate chief medical officer for Adult Ambulatory Clinics at Vanderbilt University Medical Center.

Earlier this year, VUMC implemented its own patient-facing AI tool into the patient portal. The technology was designed to help patients craft better, more detailed messages to their providers. The AI can probe patients and ask clarifying questions, but it can't dispense advice, at least not yet.

According to Aher, that's because VUMC's patients still primarily want to hear from their healthcare providers, not a chatbot, for medical advice. Armed with that knowledge, the organization designed its AI to provide value to what the end-user -- in this case, the patient -- wants.

Although Aher said that's going to be important on the provider side, as well.

"What we're seeing across healthcare is the abandonment of a large number of AI tools," he noted. "People got really excited about it, people continue to be really excited about it. They're rushing to try to get these tools and to achieve certain outcomes, but a lot of these things are not providing the value that we thought they were."

Getting ROI from an expensive AI tool will require health system leaders to know how the medical workforce will actually use AI and how it can integrate into their day-to-day work.

For example, tools that have great functionality but live in a separate system won't have the same uptake and impact as something that lives within the EHR or the patient portal.

"Starting with workflow and understanding how your tool is going to integrate into how somebody does their job day-to-day is probably the most important thing when it comes to adoption," Aher advised.

3. Differentiating from Big Tech's commercial AI

There's no denying that patients have flocked to the commercial AI products that companies like OpenAI and Google have developed. Indeed, ChatGPT and Gemini rise to the top as the most commonly used AIs for patients seeking medical information, according to a recent Pollfish and Insuranceopedia poll of 1,250 U.S. adults.

That begs the question: Why would patients use an AI from their healthcare provider when they already use other products with more name I.D.?

According to Shaul, it'll come down to user intent. Chatbots owned by Big Tech will likely continue to dominate the top of the funnel, he conceded, remaining popular for broad medical questions

"That is what I call very, very low intent educational level," Shaul said. "People will naturally go to other tools for simple knowledge-based questions, but they will defer to their health system and their provider for the actual clinical questions."

AI tools hosted by hospitals and health systems will be better at capturing the bottom-of-the-funnel, high-intent users. These are the people who might want to know when they are due for their next tetanus shot or cancer screening, or who want to learn what a new diagnosis means for them and their unique health needs. These people are more likely to act on or engage with the information they learn.

The data somewhat backs this up. The Insuranceopedia poll showed that a third of patients would never use AI for medical advice, showing that there's a sizeable portion who want to hear from their healthcare providers. A separate February 2026 survey showed that nearly three-quarters of patients ask their doctors for health information, while only 16% turned to AI.

In other words, the health system brand and patient loyalty still carry a lot of weight.

This means healthcare providers themselves need to maintain their interpersonal relationships with patients. This will enable them to have productive conversations that patients trust.

But it also means there's an opportunity for health systems to make their mark in the AI space.

With the backing of the health system brand, these tools can help streamline patient access to health information. But as organizations consider building their own AI chatbots, they must consider how their tool will fit into the larger health IT landscape and their user workflows.

Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT.