As health systems move deeper into AI adoption, leaders say the definition of return on investment is shifting from narrow cost savings to broader measures tied to strategic goals, workforce well-being and clinical impact. The next era of AI ROI, they say, will reflect how the technology transforms care delivery rather than how many minutes it saves.
Nasim Eftekhari, chief AI and analytics officer at City of Hope in Duarte, Calif., said healthcare is “moving past the era of measuring AI solely by ‘hard’ ROI, simple time and cost savings, toward a more sophisticated valuation of ‘cognitive ROI’ and velocity.” Instead of looking only at which administrative tasks can be replaced, she said health systems will increasingly measure “the invisible ledger of cost avoidance: identifying the readmission that didn’t happen, the claim denial that was preempted, and the cognitive load removed to prevent clinician burnout.”
In specialty-driven environments such as oncology, she said, speed is becoming a critical indicator of AI’s value.
“The true return is increasingly measured by velocity to value: how drastically we can compress the timeline for clinical trial feasibility or patient matching, effectively converting computational speed into clinical survival,” she told Becker’s.
City of Hope has seen that shift firsthand with its internal generative AI platform, HopeLLM. Ms. Eftekhari said the tool has “already saved thousands of hours of pajama time,” summarizing complex records and reducing the hours clinicians spend completing documentation outside of work. It also flags scheduling errors, boosts operational efficiency and scans unstructured clinical data to accelerate trial matching.
“It’s helping us match the right patient to the right trial at the right time, faster than ever before,” she said.
At Ann & Robert H. Lurie Children’s Hospital of Chicago, the focus is moving from measuring ROI at the project level to understanding systemwide value creation, said Rajiv Kolagani, chief data and AI officer.
Rather than calculating returns through traditional formulas, Lurie is taking a two-pronged approach to AI: reducing value attrition in workflows that typically drain time and efficiency, and driving value addition through personalized care, smarter navigation and predictive clinical risk management. If executed well, Mr. Kolagani said, “ROI will automatically be created through this process,” emphasizing that the health system is not “hung up on formulas and math.”
Sutter Health, based in Sacramento, Calif., is taking a similarly broad approach. Ashley Beecy, MD, the system’s chief AI officer, said health systems will increasingly evaluate AI based on whether it helps meet larger strategic objectives, not whether the technology itself is flashy or novel.
“Using it should never be about technology for technology’s sake or chasing the next big thing,” she told Becker’s. “Rather, does it make the care we provide better or easier to deliver?”
Dr. Beecy said phased implementations have helped Sutter assess AI tools more holistically, measuring not just efficiency but also workforce well-being.
“It is just as important to measure how AI creates better efficiencies, whether in administrative tasks or time,” she said. “And perhaps most impactful of all is how AI is supporting both your people and patients. Is it helping reduce burnout and mental fatigue of your providers so they can better meet the needs of patients?”
All of these factors, according to Dr. Beecy, have an accumulative effect that can be transformative for patients and people, “which can be invaluable.”
Others say health systems are beginning to treat AI as a platform investment rather than a collection of isolated use cases. At Denver Health, Daniel Kortsch, MD, associate chief of AI and digital health, said ROI is increasingly aligned with the quadruple aim.
“Rather than measuring success through minutes saved or short-term payback alone, performance is increasingly evaluated based on whether AI advances strategic goals,” he told Becker’s.
Denver Health deployed an open-source, on-premises large language model platform originally intended for text generation. It now supports multimodal outputs and agentic capabilities.
“A single platform investment is powering multiple applications without additional capital costs,” Dr. Kortsch said. He believes this shift will define the future of ROI analysis.
Memorial Hermann Health System in Houston is framing ROI around four categories of AI maturity, each with its own measurement strategy, said Chaitanya Vempati, associate vice president of AI and analytics. The first, “table stakes AI,” includes tools like productivity assistants and copilot-style drafting systems. Their value, he said, is “challenging to quantify,” so Memorial Hermann looks instead at staff satisfaction, retention and talent competitiveness.
The second category, AI-enhanced workflows, focuses on automation and optimization tools such as ambient documentation, prior authorization automation and imaging support. These show value through “workflow efficiency improvements,” Mr. Vempati told Becker’s, including quality metrics and burnout reduction.
More transformative initiatives, what he calls “new AI-imagined workflows,” include personalized treatment protocols and real-time deterioration prediction. These directly influence clinical KPIs such as mortality, recovery times and length of stay. The final category, exploratory “cool AI,” covers emerging technologies that are valuable for innovation planning even without immediate ROI.
Across all categories, Mr. Vempati said the industry faces data-sharing and integration barriers that make consistent ROI measurement difficult. He believes standardized AI monitoring frameworks, similar to HL7, will emerge in coming years. Ultimately, he said health systems will adopt “portfolio-based approaches to AI ROI measurement,” matching evaluation strategies to the type of AI investment.
“The organizations that succeed,” he said, “will be those that can effectively measure and communicate value across all four AI investment categories while building the data infrastructure necessary for comprehensive ROI analysis.”
