The pressure to adopt AI does not wait for an org chart.
Many hospitals and health systems have no data science group and no dedicated AI team, yet they are deploying the same ambient documentation, automation and clinical tools as organizations many times their size.
Shanidy Shook, vice president and CIO of Midland (Texas) Health, told Becker’s the real constraint of size is not money. It’s the cost of a mistake. Small and midsize systems can often move faster than their bigger peers, she said. They just can’t afford to be wrong.
“We don’t have the luxury of chasing shiny objects,” she said. “A misstep, especially as an early adopter of new technology, creates a burden on already lean teams and can create clinician distrust. I think the way that we overcome this is through discipline. We are focusing only on the highest-value use cases that have demonstrated real results at other facilities. We start small and prove value before scaling.”
Reid Health, a regional health system based in Richmond, Ind., has deployed 318 AI tools generating roughly $7.2 million in annual value — without a chief AI officer or a data science group, CIO Muhammad Siddiqui said.
“I do not think size is the core issue. Focus is,” Mr. Siddiqui said. “Larger systems can fund dedicated AI teams but also fund a lot of distraction.”
What Reid Health has instead is a defined path every AI decision must travel. The system built its AI Clinical and Innovation Committee as a subcommittee of its information systems steering committee, and every tool is evaluated against a structured framework aligned to the NIST AI Risk Management Framework — clinical risk, data privacy exposure, vendor accountability and measurable value, all assessed before anything goes live.
“The committee includes clinical and operational voices, not just IT. That is intentional,” Mr. Siddiqui said. “AI decisions at the bedside are not IT decisions.”
Shreveport, La.-based Willis Knighton Health goes further, said Mark Galea, senior vice president and CIO. The five-hospital system does not run proofs of concept as tests, deploys only solutions proven elsewhere and defines the value it is seeking before investigating any AI tool.
Mr. Galea, who owns the system’s enterprise AI strategy and reports to the CEO, chairs an AI governance council the system stood up in a 90-day sprint. Its membership spans compliance, providers, a COO and a chief nursing officer, with the system’s chief information security officer as co-chair. The result, he said, is no failed deployments or proofs of concept and high adoption rates for clinical alerts, ambient scribes, documentation review and coding review solutions.
“We first ask, why not Meditech and why not Microsoft?” Mr. Galea said. “If they have it or are building it, then our position is typically to wait and then deploy when the solutions are ‘prime time’ unless the ROI and breakeven clearly show adopting a short-term solution financially strengthens the enterprise as an interim solution.”
For systems without development shops, that approach makes vendors the delivery mechanism for nearly everything — a dependence Mr. Siddiqui called “the honest answer most smaller system CIOs are not saying out loud.”
“We rely on vendors heavily, and that is not a weakness if you govern it well,” he said. “The risk is not vendor dependence. The risk is vendor dependence without accountability. We track utilization, we measure outcomes, and we hold vendors to documented performance expectations. When a tool is not performing, we say so.”
Linda Stevenson, chief operations and information officer of Norwalk, Ohio-based Fisher-Titus, said the same calculus applies at her organization, where little development happens in-house. AI evaluation starts with operational leaders to confirm a tool solves a real problem, moves to the technology team for infrastructure fit and security review, and ends with the system’s information systems oversight committee — and, for larger projects, its executive leadership committee.
“Whenever possible, we do our best to align with current vendors to streamline the support and integration for the users,” Ms. Stevenson said.
At Midland Health, the reliance itself has not grown — the system has long favored buying over building, and enterprise platforms over standalone tools — but the bar for vendors has.
“What has changed is our expectation of those vendors,” Ms. Shook said. “We are asking for stronger focus on forward-thinking roadmaps, more transparency, better data protections, and outcomes we can actually measure.”
What the four organizations share is not staffing. It’s a process no AI tool gets to skip.
“What we have is a governance structure that makes every AI decision go through a defined process, and a leadership team that treats AI as an operational priority, not an IT experiment,” Mr. Siddiqui said.
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.