Augmented intelligence (AI)—also known as artificial intelligence—is already changing how health systems work—from documentation and call centers to patient outreach and predictive models. But healthcare leaders have made it clear that the next phase is not simply about adding more AI tools. It is about building the governance, data infrastructure and operational discipline needed to use those tools safely and at scale.
Across two panels at the Becker’s Annual Meeting in Chicago, leaders from Atlantic Health, Ochsner Health, Atlantic Health, Rush University System for Health and Sanford Health described a healthcare AI landscape moving quickly from experimentation to enterprise strategy. Meanwhile, the rise of shadow AI—meaning the use of unsanctioned AI tools—shows that physicians and other health professionals are looking for faster ways to solve real workflow problems.
But scaling AI safely requires more than enthusiasm. It requires approved tools, clean data, clear accountability, cybersecurity safeguards, identity management, monitoring and a willingness to redesign workflows rather than automate broken ones.
Atlantic Health, Ochsner Health, Rush University System for Health and Sanford Health are part of the AMA Health System Member Program, which provides enterprise solutions to equip leadership, physicians and care teams with resources to help drive the future of medicine.
Listen to what shadow AI reveals
In one of the sessions, Louis Jeansonne, MD, MMM, chief medical informatics officer at Ochsner Health, discussed how unsanctioned “shadow AI” tools are infiltrating clinical and operational workflows, often outside formal oversight. This has introduced challenges for governance, trust, and compliance.
People often view shadow AI as a negative because there is risk to it. But it can also be a positive, said Dr. Jeansonne.
An early example of this is people using their own bots in Zoom meetings to take notes.
“That was really a new experience for us, and it made us realize that you can't just tell people they're not allowed to use it,” he said. “It's making you aware of a need that people have. It's making their lives easier.”
Instead of viewing AI as a misconduct, health systems should treat shadow AI as a form of workflow intelligence, Dr. Jeansonne advised. At Louisiana-based Ochsner Health, leaders made the decision to provide another solution before telling people they couldn't use those tools.
It’s important to offer people approved tools and to use these situations to figure out what the needs are, he said. Ochsner also surveys people about what AI tools they're using, to keep tabs on AI usage because “it actually gives you a good idea of what people want and what people need."
That said, patient privacy continues to be a huge risk with AI, Dr. Jeansonne acknowledged.
In general, people know not to put patient names into ChatGPT or similar tools.
“However, I don't think people think as much about the fact that if you include enough detail about yourself, your patients, dates, and times, it can become identifiable,” especially if that information is combined with another dataset that may be accessible at some point, he said. “That's probably the biggest risk that I think about."
Build governance before scale
Dedicated committees for governance, trust and compliance around the use of AI in healthcare are needed in order to provide safe AI tools to meet the needs of clinicians, said Louis Jeansonne, MD, MMM, chief medical information officer at Ochsner Health. This can prevent the use of “shadow AI,” or unapproved AI tools.
“As long as you provide an effective solution, people will use it,” said Dr. Jeansonne.
He also stressed the importance of protecting patient privacy with the use of AI, noting that “education is essential so that as new tools come online, people are always thinking with a security first mindset.”
AI adoption is giving way to responsible scaling, trust and measurable impact, noted Sunil Dadlani, executive vice president, chief information and digital officer, and chief cyber security officer at Atlantic Health, which offers care in New Jersey, Pennsylvania, and the New York metropolitan area.
To do this, you need a governance structure, otherwise chaos and fragmentation may result, Dadlani said.
The goal is to work toward a “native” AI. This means “leveraging AI and building the intelligence layer in each and every facet of the organization, whether it's starting from clinical to operations to revenue cycle, patient experience and engagement,” he advised.
Atlantic Health in its own AI applications has seen measurable return on investment as far as reducing physician burnout and decreasing pajama time.
“We've also seen improvements in patient experience and engagement, more accurate clinical coding, and, over time, improvements in patient throughput,” said Dadlani.
Agentic AI, which helps patients schedule appointments, identify appropriate physicians and guide them through procedures, has improved compliance rates. An outbound AI agent for colonoscopy, for example, reduced no-shows and cancellations by 40% in 30 days. About 10% of patients not only confirmed their appointments but actively interacted with the agent, which helped build trust and engagement, he explained.
As health systems move into this new paradigm, it's important to understand that in healthcare, when technology acts, that's where value is created. But it's also where the highest risk is generated, Dadlani said. “You have to be very cognizant … and careful about what the use case is and what kind of different flavor of AI that you want to deploy.”
Organizational readiness is equally important, he emphasized. As AI advances at an unprecedented pace, it's reshaping the skills organizations need. There will always be a gap between the rate of change happening outside the organization and the rate of change happening inside it. While this is normal, the question is whether that gap remains manageable.
If the gap becomes too large, organizations risk becoming stagnant and falling behind. If they manage it well, they can successfully adopt new technologies and upskill their workforce, said Dadlani.
“The biggest challenge is—and this would be my recommendation to everybody—is never start with the marketing brochures, the algorithms, the technology, or the vendor. Start with the problem or the challenge you're trying to solve,” said Dadlani.
It’s also important to look at things from the perspective of the entire health system. Once you've identified the system-level problem, then you can determine what solution is most appropriate,” he said.
Fix the data and workflow first
AI exposes bad data, nonstandard processes and weak accountability. Health systems shouldn’t wait until everything is perfect, but they do need to fix foundational problems as they go, said Jeff Gautney, senior vice president and chief information officer at Rush University System for Health in Chicago.
This starts with an operational and agentic AI reality check. Don’t assume that AI will fix broken systems, Gautney said. Rush experienced this firsthand when it tried to expose clinic hours through an AI agent.
“What we found was that clinics updated their hours by sending an email to marketing, then sent an email to the website provider. So, the clinic hours were always two or three days behind what they actually were,” he said. The AI agent didn't care about any of that. It simply grabbed the clinic hours that existed.
What looked like an AI hallucination was bad data—and a bad process behind the data, Gautney explained. Rush has been moving forward with the technology while simultaneously going back and fixing those issues.
The one thing you can't do is say, “I'm not going to do anything until I completely clean out the attic and reorganize everything,” he said. If you take that approach, you'll be frozen in time for two years until you try to solve that problem.
Prepare for digital workers
Rush is moving toward an “agentic enterprise,” with autonomous agents doing lower-level work and supporting physicians and care teams. To do this, operations leaders must learn to manage a digital workforce, not just wait for IT to monitor everything, cautioned Gautney.
Technology is changing so quickly that solutions implemented today may be out of date within 12 to 18 months, sometimes within just a couple of cycles.
Much of this will be about turning on the lights for operations leaders, rebuilding infrastructure and figuring out how to manage the cost of these systems.
Gautney warned that organizations should be careful as the cost of implementing an agent can quickly—and sometimes unexpectedly—consume resources.
Unlike traditional IT systems, where things are binary—on or off—these systems don't fail all at once. They tend to degrade slowly, and they can recognize when they're wildly out of sync. But when they're just starting to drift, that's on operations. It represents a fundamental change in how to think about those roles, said Gautney, who also highlighted the importance of AI monitoring.
There needs to be a structured and automated way of monitoring not just return on investment, but all the other things you have to track with these models, whether it's drift, bias, or simply whether the system is doing what it's supposed to be doing, he said.
Cybersecurity is a specific concern, Gautney continued. There have been cases where vulnerabilities were uncovered in systems that were previously assumed to be secure. In some cases, organizations had been monitoring those systems for years without realizing the risks that were there.
Don’t assume that all technology partners will act as responsibly as you would like, he cautioned. You also can't rely on regulation to keep pace in a period where it’s being struck down right and left.
“Monitoring is going to be essential,” Gautney urged.
Use AI to close access gaps
Amid the concerns, AI remains a useful tool in bridging access gaps, addressing clinician shortages and supporting proactive outreach in value-based care, said Brad Reimer, chief technology and digital officer at Sanford Health, headquartered in Sioux Falls, South Dakota.
Sanford Health has implemented AI across different areas—partner solutions, internally developed predictive models, and generative automation use cases. In particular, it has seen promising results with different machine learning models, particularly around chronic kidney disease.
“In a classic scenario, we know who our high-risk patients are. But then there’s the other 90% of the population,” Reimer said. “Our machine learning models have been able to look at that 90% and identify patients we should be proactively reaching out to for additional screening.”
Through that work, Sanford Health has doubled the number of screenings, tripling early diagnosis of chronic kidney disease. There is an economic impact to this, but the bigger story is the impact on patient care and quality, especially in rural areas where chronic kidney disease, colorectal cancer, and similar conditions are more prevalent, and where physician shortages are also the most acute.
“It’s going to bridge the gap in a really meaningful way for patients,” said Reimer.
Make AI everyone’s work
Ultimately, it's operational discipline that matters most. The healthcare leaders who define the future of AI won't be the ones with the most AI tools or even the fastest adoption. They’ll be the ones “who have scaled it, who have delivered the maximum impact, and have the operational discipline to manage this life cycle,” said Dadlani.
Organizations need leadership structures, education and forward-deployed engineering models that put technologists closer to operations, said Reimer. It’s about getting people to take a step back and spend a little time reimagining and reexamining their workflows.
Gautney reinforced the shift toward people working with technology and assisting AI, rather than the AI assisting them, noting that “we’ll see more of that in diagnostic areas as well.”
All the back-office functions are also ripe for change, with changes to improve the workflow of HR departments, as well.
From AI implementation to digital health adoption and EHR usability, the AMA is fighting to make technology work for physicians, ensuring that it is an asset to doctors. That includes the AMA Center for Digital Health and AI, which works to ensure physicians help shape how AI is developed, implemented and regulated across the healthcare system, with patient safety and physician-led care remaining at the center of those efforts.
