The following is a guest article by Eric Farr, Principal and Executive Officer at BrainStorm Inc.
Health systems stuck in pilot purgatory are not waiting on better algorithms. They are waiting on a workforce that knows what to do with AI once it arrives.
At HIMSS26, a familiar pattern emerged. Health systems are no longer debating whether to adopt AI. They are debating why it is not scaling. Ambient documentation tools are live. Revenue cycle automation is running. Clinical decision support is deployed. Yet many organizations remain stuck in pilot phases, unable to move AI from controlled experiments into operational infrastructure.
The usual explanations—vendor immaturity, interoperability gaps, and budget constraints—do not fully explain it. Some of the best-resourced systems in the country are stuck alongside community hospitals with a fraction of the budget. The real bottleneck is not technology. It is people.
Most health systems deployed AI as a technology initiative and treated workforce readiness as a training checkbox. The systems making real progress treated readiness as infrastructure: continuous, role-specific, and tied to governance from the beginning.
Completion Rates Are Not Readiness
The standard approach to AI workforce preparation is familiar: buy a tool, build a training module, track completions, and report the number to leadership. It is the same model health systems have used for EHR rollouts, compliance requirements, and other enterprise technology initiatives.
For AI, that model breaks down. Capabilities change too quickly for point-in-time training to keep up. A module built in January can be outdated by March—not because the content was wrong, but because the tool has new capabilities, new risks, and new workflow implications.
More importantly, completion tells you who sat through the content. It does not tell you whether a radiologist is verifying AI-flagged findings before acting, whether a revenue cycle analyst is catching AI-generated coding errors before they become denials, or whether a clinician is using an approved AI tool instead of bypassing it for an unapproved consumer app.
That last scenario—shadow AI—is where the readiness gap becomes most dangerous. When staff do not feel confident or supported using approved tools, they find their own workarounds. In healthcare, those workarounds can involve patient data, and organizations often discover them only after something has gone wrong.
What Readiness as Infrastructure Looks Like
Organizations moving past pilots tend to share four traits.
First, readiness is continuous, not episodic. When a tool changes, staff get guidance in the flow of work—not months later in an annual refresher.
Second, readiness is role-specific, not generic. A nurse using ambient documentation faces different decisions than a coder using AI-assisted charge capture. Generic AI awareness training serves neither one well.
Third, readiness is measurable at the behavior level. Completion rates are input metrics. The better question is whether staff are using approved tools, following verification protocols, and escalating when outputs look wrong.
Fourth, readiness is auditable. Under regulatory or compliance scrutiny, organizations may need to show not just that they had a policy, but that workforce preparation translated into practice and was maintained over time. That requires evidence, not a slide deck.
The Community Hospital Problem
This issue is not limited to large academic medical centers. In some ways, it is even harder for smaller systems. A 200-bed community hospital may not have a dedicated AI governance team or a chief AI officer. It may depend on an EHR vendor’s roadmap and whatever training materials ship with the product.
But that hospital is still deploying AI into workflows that carry real regulatory and operational exposure. The compliance standard does not shrink just because the org chart does.
For smaller systems, readiness infrastructure has to be low-overhead and scalable. It has to adapt as tools change, reach people where they already work, and produce the behavioral evidence governance requires. Otherwise, organizations are left hoping a one-time training session was enough—and finding out only after a failure.
The Question That Separates Scaling from Stuck
If you want to know whether your organization is ready to move AI out of pilot, ask one question: Can you produce evidence—right now—that the people using AI in clinical and administrative workflows are demonstrably prepared, and that their preparation is being maintained as the technology changes?
If the answer is a training completion report, you are not ready. You have checked a box. The health systems scaling AI did not check a box. They built a system.
AI will continue to get faster, more capable, and more embedded in care delivery and operations. The workforce readiness gap will not close on its own. It closes when organizations decide that preparing people is as important as deploying the technology—and build the infrastructure to prove it.
Eric Farr is Principal and Executive Officer at BrainStorm Inc., which he co-founded in 2002. A Wharton MBA and 2019 EY Entrepreneur of the Year (Utah Region), he has spent more than two decades working with organizations on the gap between technology access and real workforce capability. BrainStorm helps enterprises build the human readiness infrastructure needed to capture durable value from AI and enterprise software.
No comments:
Post a Comment