As AI accelerates across healthcare, some CIOs are deliberately slowing things down.
At South Shore Health, based in South Weymouth, Mass., the challenge isn’t a lack of interest in AI — it’s managing the volume of ideas, tools and requests pouring in, while keeping the organization focused on what actually improves care. For CIOs, the work has shifted from adoption to restraint.
When Sam Ash, MD, stepped into the CIO and vice president of IT and innovation role, his first priority wasn’t launching new AI initiatives. It was, as he put it, “taking stock of where we are so we could figure out where we need to go.”
Nearly every function in modern healthcare now depends on software, Dr. Ash said, which makes it easy for problems to cascade in unexpected ways. That complexity, combined with a constant flow of requests, can push IT teams into reactive mode — fixing issues as they arise without a clear long-term direction.
To counter that, Dr. Ash focused early on governance: creating structure that allows the organization to say no as often as it says yes.
At the center of that effort is a newly formed executive council made up of clinical, financial and operational leaders. The group helps set priorities not just for the coming year, but over a multiyear horizon, acknowledging that AI capabilities will continue to evolve — and that not every promising tool belongs in production.
That top-level oversight is paired with several subcouncils focused on specific domains, including clinical applications, enterprise systems and operations. Those groups include frontline experts such as application leaders, whose day-to-day experience often reveals whether an AI request solves a real problem or simply adds noise.
This structure has become especially important as interest in AI accelerates.
Dr. Ash said return on investment looks different depending on the use case. Some tools, such as productivity assistants, produce benefits that are difficult to quantify. Others, including ambient documentation pilots, are evaluated through measurable indicators like clinician documentation time, after-hours work and potential impacts on billing and charge capture.
For clinical AI tools designed to predict patient deterioration, success may be defined by shifts in downstream outcomes — such as ICU transfers or rapid response patterns — rather than by a single headline metric. Establishing those measures in advance is now a required part of the organization’s AI evaluation process.
However, reining in AI isn’t just about metrics, it’s also about trust.
Dr. Ash continues to practice as a pulmonologist and intensivist, spending time in the ICU each week. That ongoing clinical work, he said, helps anchor technology decisions in the realities of bedside care and gives clinicians confidence that IT leadership understands the consequences of poorly designed tools.
That perspective informs how new tools are evaluated. Clinicians are brought into the process early, including direct sessions with vendors, so they can ask questions and understand how products are being assessed. When changes are unavoidable, the focus shifts to transparency and presence on the units.
A key bridge between IT and clinicians is the health system’s clinical informatics team, made up of nursing informaticists who regularly round in inpatient and ambulatory settings. Their role, Dr. Ash said, is to surface issues early and identify where technology is helping — or hindering — care delivery.
Independence adds another layer of complexity. While a single-hospital system can move quickly, it must still support many of the same technologies as larger systems with far fewer people. In that environment, Dr. Ash said, unchecked AI adoption can quickly become unsustainable.
What has helped most is collaboration with peers nationwide. Dr. Ash emphasized that CIOs are not competing on AI. “Everyone’s just trying to, at the end of the day, make sure that the patients are cared for,” he said, adding that colleagues across the country have been generous in sharing lessons about governance and structure.
As AI tools become easier to deploy and harder to resist, Dr. Ash’s approach reflects leadership today is less about how fast AI can be rolled out — and more about deciding where restraint is necessary.
