Most health systems talk about accelerating AI adoption. Sudipto Srivastava, chief data and analytics officer at Montefiore Health System in New York City, is thinking about the opposite problem.
Mr. Srivastava, who oversees data analytics, organizational reporting, AI governance, and research IT for the 13-hospital academic medical center, says the instinct to keep adding technology without equal discipline around removing it is one of the most consequential mistakes health systems can make. Montefiore has already launched more than 80 AI solutions and nearly 100 more in the pipeline; adding any more would challenge capacity.
“I like to say that we’re really good at addition, but not good at subtraction,” he said. “We’re really good at adding more and more solutions and tech to it, but we’re not as deliberate about subtracting it and taking them away.”
That philosophy sits at the core of Montefiore’s AI governance framework, which has itself gone through deliberate iteration. Mr. Srivastava said the system’s governance 1.0 was designed for speed as a response to the volume of incoming AI requests. Governance 2.0, launched at the start of this year, expanded the table to include clinical, nursing, finance, legal, privacy, and cybersecurity stakeholders for a more thoughtful and deliberate process. Governance 3.0 is already in development.
The framework is about more than approving solutions; it focuses on holding the team accountable for results after launch. When an AI tool clears governance review, it’s because the tools will touch a defined number of encounters, improve productivity to a measurable degree, or hit a specific cost or volume threshold. The team denies tools that don’t meet this criteria.
“When you launch a solution and AI governance approves it, the hypothesis was it was going to do x, you know, x number of patient encounters and improve productivity or maybe some dollar threshold or numbers or volume thresholds,” said Mr. Srivastava. “Is it doing that? And let’s be very honest to measure if so. If it isn’t, let’s take it out of our ecosystem. Trust me, these are not easy conversations to have. But that is a framework that is needed so, two years, three years down the line, we’re not taking on so much tech debt that we have to have a separate exercise to unwind all of that.”
The rigor is especially important given the breadth of what Montefiore is pursuing. Active areas span radiology, acute care, cardiology, GI, IT security and nursing, along with significant investments in revenue cycle and supply chain. On the revenue cycle side, Mr. Srivastava sees AI as a meaningful lever for denials management, prior authorization, and throughput improvement. Supply chain, which he describes as the second or third largest cost center for most health systems, is another priority area he believes is underexplored.
“They’re buying pencils to MRI machines,” he said. “Using AI in a thoughtful manner is huge.”
Population health and care management round out the clinical priorities, with work already underway on care gap closure, HEDIS measures, and cancer auto-tagging using BioClinical BERT. On the research side, Montefiore entered a partnership last year with Dandelion Health, joining a consortium of health systems offering de-identified data for AI algorithm development and advanced research.
But resources aren’t unlimited.
The macro environment heading into 2026 and 2027 will force renewed attention on cost savings. Data protection remains a persistent concern, particularly the questions that arise around vendor data use and LLM training agreements.
“Healthcare is embracing AI at a faster pace,” he said. “The interest is there at the leadership level, the clinical level, the nursing level. Then how do you filter out the signal from the noise? What is worth investing in? What is not worth investing in?”
Staying grounded in that environment requires a structural choice Montefiore made deliberately: rather than appointing a separate AI czar, the organization embedded AI responsibility across every role.
“Our philosophy is that AI and tech is everyone’s responsibility,” he said. “We find that this frees up our staff and doctors to explore things from their unique vantage point.”
The change management associated with new technology and AI tools is critical for maximizing return. Health systems can’t force adoption and the anxiety clinicians and staff feel about new tools is taken seriously rather than dismissed.
“There has to be a conversation and understanding of what it does, what it doesn’t do,” Mr. Srivastava said. “There’s a lot of training and coaching that goes in there in terms of what is the value of that, talking and being actually honest about some of the skepticism that people have. That is a journey. Sometimes you take two steps forward and one step backward.”
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
