Health system CIOs, revenue cycle leaders and chief innovation officers gathered at Becker’s 11th Annual IT + Revenue Cycle Conference to share what is working — and what is not — as AI reshapes every corner of health system operations.
Across five sessions on Sept. 14, a clear picture emerged: The technology is moving faster than most organizations can absorb it, and the edge will belong to systems that pair strong governance with genuine clinical and financial discipline.
Here are 10 trends driving the conversation:
1. AI is everywhere — and governance is the differentiator.
Several health system leaders at the conference described being overwhelmed by the volume and velocity of AI tools arriving at their doorstep. The leaders pulling ahead are not those with the most tools; they are the ones with the clearest governance structures to evaluate, monitor, and sunset them. Doug King, CIO of Chicago-based Northwestern Medicine, described a framework built around five core value metrics — ROI, patient experience, quality, patient safety, and revenue — and a willingness to kill pilots that do not deliver.
“We agree upfront what the value is we’re looking for,” Mr. King said. “If it doesn’t meet it, it needs to be shut down. We can always revisit it, but for that pilot, there are 10 more waiting in line.”
2. AI voice agents are scaling across patient outreach, and patients are responding.
Charlotte, N.C.-based Advocate Health has deployed an AI voice agent named Anna across more than a dozen use cases — from hypertension follow-up to colonoscopy prep to medication management — and the results have surprised even internal champions. Bobbie Byrne, MD, Advocate’s EVP and CIO, noted that patient refusal rates are low and that some calls run more than 20 minutes, something no human staff member could replicate at scale.
“There is no human who would sit and do a 22-minute hypertension follow-up call,” Byrne said. “Once people realized what we were able to do, they started thinking: How can I use that for my untapped issue?”
3. Workflow redesign — not just technology layering — is the real imperative.
Multiple speakers warned against bolting AI onto broken processes. Philadelphia-based Penn Medicine’s Mitchell Schnall, MD, PhD, described a fully ambient, keyboard-free clinic scheduled to open within the year — not as a technology showcase, but as a wholesale reimagining of the care workflow. Houston-based Memorial Hermann’s chief digital officer, Eric Smith, echoed the point, describing a voice AI initiative that stalled because the underlying process had not been examined first.
“We tried to layer cool technology on top of a process that probably needed to be relooked,” Mr. Smith said. “It really wasn’t a technology failure. It was us being introspective.”
4. Revenue cycle is moving from transactional to strategic, and the workforce must evolve with it.
Drew von Eschenbach, VP of Enterprise Revenue Cycle for Seattle-based UW Medicine, described a deliberate shift away from siloed, heads-down billing work toward a smaller, more analytically sophisticated team. At a $6 billion system, the FTE count in revenue cycle has remained flat even as volume has grown — a direct result of automation absorbing transactional work while higher-caliber staff handle exceptions and strategy.
“Our job in revenue cycle is to convert the care that we deliver into cash so we can continue the investment and delivery of improving healthcare,” von Eschenbach said. “Every dollar that goes into revenue cycle is one less dollar that goes into patient care.”
5. Cybersecurity is becoming the defining risk of the AI era.
When asked what topic would dominate next year’s IT conference, Dr. Byrne did not hesitate: cybersecurity. As AI accelerates the speed at which threat actors can probe and penetrate systems, the concept of a zero-day vulnerability is giving way to zero-hour exposure. She and her peers are watching agentic AI be used offensively and racing to use it defensively.
“Zero day is not going to be zero day — it’s going to be zero hour,” Dr. Byrne said, describing a future when AI agents conduct attacks and defenses in near real time.
6. Virtual and asynchronous care is shifting from experiment to expectation.
Advocate Health is conducting roughly 500,000 virtual primary care visits annually through a model where a patient’s named physician operates exclusively in a virtual environment. Aurora, Colo-based UCHealth has launched virtual-first primary care — where virtual is not a supplement but the entire care relationship — and reports strong uptake, better care gap closure, and higher patient engagement among segments that prefer asynchronous communication.
“There is a large segment of our patients who prefer asynchronous, who prefer virtual,” said Richard Zane, MD, chief medical officer and chief innovation officer at UCHealth. “Follow-up is better. We’re able to close care gaps easier. We’re able to communicate with certain segments of the population better.”
7. Ambient AI is changing the clinician experience but requires careful implementation.
Ambient documentation is driving measurable improvements in physician satisfaction and efficiency. Dave Newman, MD, chief medical officer of virtual care at Sioux Falls, S.D.-based Sanford Health, described a patient who told him after a decade of care: “You’ve never looked me in the eyes.” After implementing ambient AI, that changed. Sanford providers using ambient technology are leaving work seven minutes earlier per visit and scoring higher on patient satisfaction.
But Dr. Byrne cautioned that nursing implementation is more complex, requiring workflow and culture change before the technology can succeed.
“We had to pull back on the voice implementation and put in pre-training we call ‘care out loud,’ helping nurses get more comfortable speaking their care as they do it,” she said.
8. AI is producing measurable ROI in revenue cycle — particularly in prior authorization and denial management.
Penn Medicine estimates more than $100 million in AI-driven performance gains over the next few years, with meaningful returns already visible in prior authorization automation — a 30% efficiency improvement — and denial management. Springhill Medical Center CEO Diana Allen, PhD, described using AI to mine payer rejection data at scale, identifying a $10 million NDC problem across payers in 90 days that the old queue-based model would have missed entirely.
“When you use the tool and let the data reveal itself, that is the tool,” Dr. Allen said.
9. The CIO role is becoming an enterprise leadership role — not a technology role.
Philadelphia-based Jefferson Health’s Luis Taveras, PhD, described a CIO’s job as general management that happens to specialize in technology — and warned that leaders who stay in the technology lane will be left behind.
Donna Roach, CIO at Salt Lake City-based University of Utah Health, described spending nearly a year building a federated IT governance structure that brought academic departments, clinical operations, and health system IT into a unified model. Both agreed the most successful CIOs of the next era will be defined by their executive relationships, not their technical depth.
“The most successful CIO will be the one who has the relationships with their executive peers,” she said. “Do you have their cellphone number, and when you text them, do they respond?”
10. Access, scheduling and patient-facing AI are the next frontier — and the bar is being set by other industries.
Dr. Zane put it plainly: The health system is in the top 1% of systems Epic serves for online scheduling — and that is still not good enough. Patients arrive at appointments having already consulted ChatGPT, Claude or other AI tools, and they increasingly expect healthcare to match the seamlessness of airline booking or Amazon.
Children’s Hospital Los Angeles’ Omkar Kulkarni described the next evolution: patients building their own AI agents to navigate insurance authorizations, specialist referrals, and care coordination across multiple providers.
“AI is just math; it is not magic,” Dr. Zane said. “We have to take care of patients where they are, where they live, in their home if we can, when they want it — not when we want it.”
