AI's Last Mile in Healthcare Runs Through a 12-Hour Shift

Most conversations about AI in healthcare focus on the technology. Which models are accurate enough. Which vendors integrate with the electronic health record. Which use cases pay back fastest.

Those questions matter. But the hardest part of AI in a health system usually isn’t technical. It’s the last mile: the moment a tool meets a nurse, a resident or a respiratory therapist halfway through a 12-hour shift, with three patients waiting and no time to figure out something new.

That’s where adoption is won or lost. And it’s a leadership problem far more than a technology one.

Using it without being led into it

In July, Incredible Health published its 2026 State of Nursing Report, based on a national survey of 2,240 nurses. In a single year, the share of nurses using AI nearly tripled, from 15% to 44%.

That sounds like progress, and in part it is. But look at what came with it. Only 8% of nurses said their employer had set a clear AI strategy for their role. Forty-six percent had received no AI training in the past year. Nearly half of those using AI said it saved them little or no time, and nurses who had been trained were more likely to report real time savings than those who hadn’t.

In other words, the workforce is moving faster than the leadership around it. People are picking up tools on their own, without a clear sense of what they’re for, where the guardrails are or how their work is supposed to change. That’s not a technology gap. It’s a gap in direction.

What the last mile actually requires

When a new tool arrives on a unit, people quietly ask the same questions. Will this help me or watch me? What happens if I don’t trust its recommendation? Who decided we needed this, and did anyone ask what we actually struggle with? Is this the first step toward needing fewer of us?

No vendor can answer those questions. Only leaders can, and the leaders closest to the work are usually the ones who haven’t been equipped to.

That’s why so many promising deployments stall after the pilot. The technology performs in the demo. Then it reaches a unit where the manager wasn’t part of the decision, the staff weren’t trained and no one explained how the tool fits into the way that team actually works. The tool gets used inconsistently, the results look disappointing and the organization concludes the technology wasn’t ready.

Sometimes it wasn’t. Often the people around it hadn’t been prepared.

There’s a trust cost too. Every tool that arrives without explanation teaches clinicians that change is something that happens to them. The next tool, even a better one, then starts from a deficit.

Leading the last mile

Start with the work, not the tool. Before deploying anything, ask the people doing the job where their time goes and what frustrates them most. AI that solves a problem a team has named gets adopted. AI that solves a problem leadership assumed gets worked around.

Tell people what it means for their role. A clear, honest explanation of how a tool changes a job, and what it doesn’t change, does more for adoption than any feature. If only 8% of nurses have that clarity, every system has room to stand out.

Train in the flow of the shift. Training that happens in a classroom months before go-live rarely survives contact with a busy unit. Short, practical guidance from a trusted colleague at the moment of use works far better.

Protect judgment. Make it explicit that clinicians are expected to question an AI recommendation when it doesn’t fit the patient in front of them, and that doing so is valued, not penalized. People bring their best thinking when they believe it’s wanted.

Equip the managers. The unit manager decides whether a new tool becomes part of the team’s rhythm or another login nobody uses. Give managers the context, the authority to adapt and a direct line for feedback.

I explored this from the workforce side in Leading Teams When AI Does More of the Work. In healthcare, the stakes are higher, because the last mile ends at a patient.

The people side comes first

Health systems are right to invest in AI. The opportunity to give clinicians time back and improve care is real. But the systems that see the returns will be the ones that treat adoption as a people strategy. As I wrote in Healthcare Needs People Experts Now, the strategy that matters most eventually runs through the people who have to live it.

The technology may shine in the demo. The last mile is still a 12-hour shift. Lead it accordingly.

Want to explore these ideas further? Learn more about my work on leadership, identity and conviction at www.theglennllopis.com.