Artificial intelligence has entered a new phase inside health systems.
For the past several years, AI initiatives largely lived in innovation centers, small pilots and controlled vendor tests. Today, at organizations such as Houston Methodist, MultiCare Health System and Mount Sinai Health System, AI is becoming embedded in workforce strategy, operational design and executive governance.
The shift is being driven by necessity, not novelty. Leaders are confronting clinician burnout, documentation overload, tight margins and growing access demands. AI is increasingly positioned as a tool to address all four.
Burnout data changes the equation
At Tacoma, Wash.-based MultiCare Health System, ambient clinical documentation was initially evaluated like any other technology purchase. But as more companies hit the market with advanced ambient listening tools, Michael Han, MD, vice president and chief medical information officer, decided to evolve his approach. He led a head-to-head comparison of three vendors, enrolling roughly 550 physicians and advanced practice providers in what he referred to as a “Bake Off.”
The team measured time in chart, after-hours work, productivity and coding outcomes. But the most significant finding was tied to workforce well-being.
“Prior to using an ambient clinical documentation tool, 60% of our physicians and APPs, experienced at least one symptom of burnout,” Dr. Han said. “After using the ambient clinical documentation tool, 16% of our physicians and APPs using the tool were experiencing at least one symptom of burnout. We were able to reduce symptoms of burnout by 75%, which was astounding.”
The results extended beyond burnout symptoms. Prior to trialing the ambient listening technology, 11% of physicians and APPs said they enjoyed their work; after the trial, 37% reported the same agnostic of vendors. The findings reframed the conversation around AI from a productivity enhancement to retention strategy.
“If you’re not at least looking at an ambient clinical documentation tool for your physicians and APPs, you don’t care about your physicians and APPs. And I’m quite serious,” he said.
As MultiCare scales its chosen solution, the next challenge is governance. Dr. Han said AI oversight will be a top priority in 2026, with a goal of developing what he described as a “single pane of glass” to monitor tools for bias, safety, drift and return on investment across the enterprise.
Moving beyond the ‘coolness factor’ to real value
At Houston Methodist, leaders have experienced the rapid expansion of AI vendors firsthand. Michelle Stansbury, associate chief innovation officer and vice president of IT applications, said the organization previously ran dozens of pilots simultaneously.
“For a while, we had probably 25 pilots going on at a time,” she said.
That approach has shifted toward discipline and maturation. Rather than layering on niche solutions, Houston Methodist is focusing on enterprise partners and expanding proven use cases, including ambient listening beyond physicians to hospitalists and now nurses.
“AI is just exploding across the healthcare industry and I think all over the place. You really have to decide what you are going to do. What are those things you really want to focus on that are really going to bring the greatest value to the organization?” Ms. Stansbury said.
Houston Methodist’s strategy reflects mounting financial pressure across the industry. Every technology investment must demonstrate measurable value.
“You can really get caught up in the coolness factor,” she said. “But what are the models? What are the things that they have built on top of the AI? And how is our data? Is our data clean enough for us to be able to use and trust it? Because AI is built overall to increase efficiencies, and that’s what it should be. It’s not right overall to add cost to the organization, or add processes, and not be able to reduce overall the labor cost with some of these efficiencies.”
Instead, Houston Methodist is targeting documentation burden across care teams. Ms. Stansbury and her team spent around 18 months examining ambient listening solutions for nurses and talked to several vendors, but none gained traction until Epic released it’s ambient listening solution for nurses.
“We’re getting ready to start a pilot with them in our Cypress hospital,” said Ms. Stansbury. “I’m very anxious to see how it works because nurses are just as overburdened with documentation as physicians are, and whatever we can do to help that overall, we will. We started virtual visits to be able to take admits and discharges from them and take on overall rounding activities, but there’s still that documentation burden.”
Houston Methodist aims to create an intelligent healthcare system of the future powered by technology and AI. But for the next year they’re focused on which use cases will make the biggest impact right away for the clinical teams.
“We’re really trying to look at what is bringing the greatest value to the organization,” Ms. Stansbury said. “We’re all being hit hard right now from governmental changes to payer changes and everything else, and yet you want to provide the best care overall to your patients and having our clinicians utilize the best tools will help them do that. But we still have to remain financially stable.”
Governance as infrastructure
At Mount Sinai Health System in New York City, the focus over the past year has centered on building the structures necessary to scale AI safely. Girish N. Nadkarni, MD, chair of the Windreich Department of Artificial Intelligence and Human Health and chief AI officer, said the system designed governance structures to streamline decision-making while maintaining oversight.
“We sort of harmonize and streamline all of the government structures in order to make sure that it becomes more of an enabler and a force multiplier rather than a blocker,” he said.
Mount Sinai divided AI oversight by persona — patient-facing, clinician-facing, workforce-facing and researcher-facing — with executive-level escalation for high-risk decisions. The system also established what Dr. Nadkarni described as an AI assurance lab designed to make sure that AI is safe, effective, responsible so that anything that gets deployed into clinical care is assured and eliminates bias.
Alongside governance, Mount Sinai is expanding AI’s role in access and care redesign. Nicholas Gavin, MD, vice president and chief clinical innovation officer, said asynchronous care models are part of a broader effort to reduce dependence on traditional visit structures.
“We don’t want to be dependent on a model where a patient has to find time to get on a video visit or go to a clinic in order to receive care,” Dr. Gavin said.
Generative AI is central to that strategy.
“Patients with the chatbot provide a unique opportunity in order to augment care and improve access and provide shift care from reactive to proactive at a scale that has never been done before,” Dr. Nadkarni said.
At the same time, Mount Sinai’s leaders are cautious about overextension.
“We are being pitched different solutions every single day,” Dr. Gavin said. “I think it’s really, really important to focus on execution and focus on execution around core problems for the organization as opposed to getting distracted by shiny objects in the periphery.”
Dr. Nadkarni emphasized the importance of clarity and patient-centeredness. “It’s important to start with the why we are doing this and also keep the patients at the center of all of this and figure out what’s the best, most scalable way to help them safely and effectively,” he said.
