John Orsini has a simple test for the AI tools Chicago-based Northwestern Medicine buys: Prove the savings or get turned off.
“Once it’s in, it can’t be the roach motel,” Mr. Orsini, executive vice president and CFO of the system, said during a panel at Becker’s 14th Annual CEO + CFO Roundtable.
Mr. Orsini said supply chain may be the fastest place for health systems to show a return on AI, but only if leaders are disciplined about it. “AI is not cheap or free,” he said. “So if it costs X, where’s the money being saved? How many fewer people are we going to use? You have to be very deliberate or you’re going to find yourself in even bigger trouble.”
Here are the rules he applies at Northwestern:
1. Results are due in year 2. Most tools take about a year to install, so Mr. Orsini expects “meaningful results compared to what we projected” in the second year. Implementations also have to stay on schedule and on budget. “If we’re not seeing it in the first year go live, then we’re going to start to question it,” he said.
2. Tools that underdeliver get shut off. Northwestern has turned off tools a couple of times, Mr. Orsini said. People were not happy about it, but the system cannot keep adding to its cost structure.
3. Safety wins, then ROI. When a supply chain proposal competes with a clinical one, a project that improves patient safety or reduces risk “wins out every time,” Mr. Orsini said. After that, the best idea with the best return gets funded. He described a roughly $1 million app that would cut the time nurses spend abstracting hospital-acquired infection data from eight hours to two. He is willing to make the investment, but if the app costs $250,000 to $300,000 a year to run, he expects the team to show savings that cover it. Otherwise, he said, they can keep doing the work the way they do now. “If you don’t do that, all this stuff will get added, your operating margin will go down, and then they’ll be looking for a new CFO at some point.”
4. No new applications for now. Northwestern has frozen new apps. Mr. Orsini said software maintenance costs pile up, vendors of point solutions keep cross-selling each other and much of the spending is duplicative.
5. Clean data comes first. Before putting AI in front of staff, Northwestern worked down its matched exceptions, or mismatches between requisitions, receiving documents and invoices. Those exceptions are now minimal, Mr. Orsini said. An accounts payable portal that lets vendors take a discount for faster payment has turned the department into “a profit center.”
6. Start with the people who want it. “Attach yourself to that person and go as far and fast as you can,” Mr. Orsini said. Bringing a tool to a group that is not ready means it “dies on the vine.”
Next up for Northwestern is pharmacy, which Mr. Orsini called the system’s “next frontier.” Inventory is not tracked perpetually across all units, and the rebate process is “quasi cash based.” Northwestern plans to use AI to tighten both processes. The system is also working with Clarium, the panel’s sponsor, to update physician preference cards.
Amanda Chawla, chief supply chain officer at Renton, Wash.-based Providence, offered a benchmark on the same panel. During her time at Palo Alto, Calif.-based Stanford Health Care, the system implemented 26 AI use cases over about two years. Those use cases saved roughly 17,000 hours and $8 million to $10 million in hard-dollar savings, she said.
