Healthcare's AI push is creating a new workforce challenge

Across the healthcare organizations we work with, there is a lot of excitement about AI right now. Leaders see it as a way to take some of the administrative weight off their teams, move workflows along faster, and give clinicians better information at the point of care. That excitement is warranted. In a recent provider survey, 52% said they are confident AI will help simplify and accelerate administrative processes, and 87% believe technology advances will improve health outcomes over the next five years.

But here is the part I keep telling clients: excitement does not automatically turn into long-term outcomes. And right now, with reimbursement pressure where it is, health systems cannot afford to invest in AI and not see those outcomes come back to them.

We are already seeing the gap. A Black Book Research survey found that 70% of hospital leaders had at least one AI pilot that never scaled past a limited deployment, usually because the endpoints were not well chosen, the data was not there, or the tool did not actually fit the clinical workflow. That tracks with what I hear in the field. Organizations are moving fast, sometimes faster than their governance can keep up with.

The work behind every deployment

The technology itself is only part of the story. Behind every AI rollout is a growing amount of operational work that often goes unnoticed. Executives see the demos and the productivity gains. The IT and operations teams are living with a much more complicated reality.

Every deployment raises a familiar set of questions. Which data sources are being accessed? How are outputs being validated? Who is watching for inaccurate recommendations or workflow disruption? What safeguards protect patient information? How are users being trained and supported once the tool is live?

In healthcare, these questions carry more weight than they do in other industries. A confusing or incorrect output can affect patient safety, clinician trust, compliance, or reimbursement accuracy. AI also does not behave like the enterprise software our IT teams are used to. Outputs shift depending on data quality, workflow design, user behavior, and vendor updates. You cannot just turn it on and walk away.

AI requires continuous operational oversight

What I am hearing from clients is that AI creates ongoing responsibilities across the enterprise. Security must evaluate privacy risk and vendor safeguards. Data teams must keep information clean and interoperable. Clinical informatics needs to evaluate workflow impact. The help desk has to support users navigating new processes and exceptions.

The challenge is that most healthcare IT departments were never structured for that kind of continuous, cross-functional oversight. The traditional model is organized around defined systems: infrastructure, analytics, cybersecurity, the electronic health record. AI cuts across all of them at once.

Another Black Book survey makes the point. Only 29% of hospitals have implemented and enforced AI policies covering model inventory, lineage, and signoffs. Just 22% are highly confident they could produce a complete, auditable AI explanation within 30 days for a regulator or payer.

Without clear ownership, you get bottlenecks. A pilot launches without standardized evaluation criteria. A tool goes live before escalation procedures are in place. Departments adopt new capabilities faster than the governance process can mature. In most cases, organizations are underestimating how much coordination is required after go-live. The initial deployment is just the start of the operational effort.

This is why I encourage leaders to treat AI adoption as an operating model question, not just an IT staffing question. The answer is not piling more work on the teams you already have. It is building governance and workflows designed to support AI, at scale.

Zack Tisch is partner of portfolio services at Pivot Point Consulting, a healthcare IT consulting firm.