How Penn Medicine is driving adoption and ensuring tech “floats to the background”

Key takeaways

Penn Medicine cut documentation time by 20 percent and after-hours work by 30 percent with AI scribes, but its leaders say lasting adoption depends on practice transformation more than on the technology. The health system piloted ambient tools in a cardiology practice, coached skeptics in person and enlisted clinician informaticists to bridge clinical and IT teams. Those lessons now shape its clinic of the future, slated to open in November 2027, where the goal is exam rooms without keyboards.

When AI-powered tools are deployed in healthcare, a lot of emphasis is placed on early wins, and understandably so. At Penn Medicine, the implementation of AI scribes has resulted in a 20 percent reduction in documentation time and a 30 percent decrease in after-hours work.

They’re the type of statistics healthcare leaders love to see. But those metrics don’t always translate into sustained AI adoption. Achieving that requires a “boots-on-the-ground” strategy focused on identifying and removing barriers, according to Mitchell Schnall, MD, PhD, SVP for Data and Technology Solutions, and Srinath Adusumalli, MD, VP and Chief Health Information Officer.

Recently, they spoke with Digital Health Insights about the strategy they’re employing to deploy AI tools, the biggest hurdles they’ve overcome and how they’re leveraging this knowledge to design the clinic of the future.

Reducing cognitive load

One of the key benefits of ambient listening is the potential to reduce cognitive load for physicians, something Adusumalli believes can’t be overstated. “Even if the time savings are modest, the relief felt from not having to write the note and not having to carry that with you after the encounter is tremendous,” he said.

Other capabilities that have scored high with clinicians are pre-charting and coding. With automated chart preparation, medical history, labs, medications and care gaps are all pulled together before a patient visit. Backend coding has proven a satisfier and has even helped some users realize they were under-coding, said Schnall.

With ambient AI in the picture, technology is being trained to help make difficult decisions with codes, which has come as a “huge relief” to some physicians, he added.

Data-driven approach

A critical factor in the positive feedback Penn Medicine has received is the deliberate, “data-driven approach” to rolling out AI tools. Piloting in a small, real-world cohort, in their case a cardiology practice, enabled leadership to accurately gauge what was and wasn’t working and make the necessary adjustments before scaling.

“We made sure everything was working and that we were starting to see the outcomes we anticipated, and then we made the technology more widely available,” stated Adusumalli. Rather than observing from afar, the team provided hands-on education and support. They asked skeptics targeted questions about what they didn’t like and offered demonstrations to help them better navigate the system. “We had to learn what’s useful for each individual so we could improve their personal experience,” said Adusumalli.

It’s all part of the “boots-on-the-ground” approach that they believe is key to driving adoption.

“A lot of people talk about ambient as one of these technologies where people pull it out of your hands,” Adusumalli noted. And while that may be true in the beginning, or what he termed the “left side of the bell curve,” the right side is a different animal. During that phase, “you have to sit with people. You have to help them think about how to take this template they’ve used for 20 years, which they think is just fine, and convince them to change.”

That, of course, involves reworking the template and workflow. It’s not an easy sell, but the results have been compelling. “At the end of the investment, the vast majority of folks tell us that it was worth it,” he said.

Utilize, not just adopt

Reaching that point, however, requires a strong focus not just on the technology but on people and process, which are often the most challenging parts, according to Adusumalli. “At this point, the technical aspect of implementation is fairly straightforward. The harder part is practice transformation,” he said, particularly in team-based care environments.

For clinicians, it means changing how they take notes to better accommodate ambient workflow. Others might have to incorporate tools into the chart prep process. “It’s a whole scale practice transformation,” he said, a factor that is often overlooked. “You need to get the whole spectrum of the bell curve to adopt, and not just adopt, but utilize across as many encounters as possible.”

The ultimate goal with ambient technology is to “connect the arc” of the patient visit, from pre-visit preparation to interaction to post-visit, to improve the experience for patients, caregivers and providers.

Making tech work for clinicians

Achieving that requires input and collaboration from those who will be using the tools, said Adusumalli, who has been closely involved with Penn Medicine’s ambient journey from the start. “We’re not trying to have clinicians work for technology. The technology needs to work for them.”

To that end, they’ve created cohorts of clinicians to act as health informaticists and provide feedback on what is and isn’t working. These teams, he noted, will serve as bridges between clinical and IT, helping to determine how technology can be leveraged in ways that are “helpful, useful, and safe in clinical care.”

Schnall, a practicing radiologist, believes informatics leaders can play a vital role in that process, which is why he has shifted some of the strategic decision-making in their direction. “These are people who know the technology and the work,” he said. “They have insights on how to make the tech work for the teams as opposed to the teams working for the technology.”

“No keyboards” policy

All of the learnings that Penn Medicine has accumulated, particularly around advanced technology, are being put into practice at the “clinic of the future,” slated to open in November 2027.

The concept originated during the process of remodeling existing clinics and designing new facilities, when leadership realized there was an opportunity to revisit patient flow and determine whether it’s designed appropriately and whether the processes make sense.

With that mindset, they prioritized virtual registration and wayfinding, among other patient-friendly capabilities. Rooms were designed to facilitate better interactions between patients and clinicians, with the recognition that ambient technology will be utilized. The idea, according to Adusumalli, is for technology to “float to the background” and let face-to-face interactions take center stage.

Taking it a step further, Schnall has issued a challenge to his team. “I don’t want a keyboard in the room,” he said. “I want anything we do to be voice-activated or touchscreen.”

And while that may not be possible right away, one thing both Adusumalli and Schnall have learned is that it’s important to aim high, especially when it comes to improving care.

When leveraged effectively, technology “should elevate and highlight the human experience for our teams and our patients,” said Adusumalli.

Kate Gamble is an award-winning healthcare journalist and editor with more than 20 years of experience covering healthcare IT. She has interviewed hundreds of healthcare leaders, and has focused on numerous topics, including digital transformation, patient experience, rural healthcare, physician and nursing leadership, and more. She can be reached at [email protected].