The healthcare AI era has given health systems too many experiments. Moving AI from pilot to practice has become a defining challenge for digital leaders in 2026, with most organizations still working to close the gap between promising local deployments and enterprise-ready tools.
At Rochester, Minn.-based Mayo Clinic, Nari Gopala has a word for what that accumulation looks like from the inside: ‘pilothosis.’
“Just like any other healthcare organization out there, we went through our phase of pilothosis,” Mr. Gopala, Mayo’s chief digital officer, said during a “Becker’s Healthcare Podcast” interview. “We had thousands of pilots running all over the organization.”
The volume was not a management failure, but a predictable outcome of how Mayo operates. The system has long run a deliberately decentralized innovation model, built on the conviction that meaningful advances in specialty care emerge from within each discipline rather than from the top down.
“Because of the multispecialty nature of our practice, we also believe true innovation often happens very close to that specialty or subspecialty, not at an enterprise level,” Mr. Gopala said.
But decentralization does not produce on its own path to scale. A strong pilot that succeeds in one department still has to travel the full distance to enterprise deployment, and without structure, that journey often stalls.
Mayo Clinic decided to take a new organizational approach to innovation. At a senior leadership level, the system designated a dedicated leader to serve as the bridge between innovation and enterprise adoption, responsible for walking AI solutions through implementation rigor, clinical governance and regulatory review. The threshold questions became consistent across every candidate: Does this tool work beyond a pilot? Can it scale to the enterprise? Is it safe for everyone to use across a multispecialty practice?
“We have now had over 300 AI solutions in practice that have already been vetted, passed through that, and primed for scale across the enterprise, while we still have a lot of other decentralized innovations going on,” Mr. Gopala said.
What makes the governance structure work is the platform beneath it. About seven years ago, Mayo Clinic introduced an internal platform designed as a chassis for AI deployment, a foundation with defined rules for how any solution connects to the broader enterprise. Reaching that scale requires plugging into the platform, not standing up another point solution.
“We will give you the chassis,” Mr. Gopala said. “We give you the rule set in terms of how you can plug in.”
That architecture changes when the scalability question enters the room. Rather than discovering after a pilot succeeds that a tool cannot integrate with the rest of the organization, teams confront that problem before investment deepens. The model also serves as a hedge against what Mr. Gopala sees as a growing industry risk. As AI models grow more capable and more accessible, the pace of tool creation is beginning to outpace the clarity of purpose behind it.
“You are going to get a lot more solutions looking for problems than solutions that specifically address a problem that can then be scaled across an enterprise and then across a large body of patients,” he said.
