The AI Prophecy vs the Ai Airport Cafe
I love people watching, and airports are a perfect space and excellent laboratories of human behavior. We’re tired, stressed, hungry, and operating on partial information, and that’s before you even get through the security obstacle course. It is basically the same conditions that define much of our healthcare system.
On a recent trip, I witnessed a scene in an airport food outlet that captures why technology alone doesn’t change behavior and, importantly, why the promise of Agentic AI in healthcare may hinge more on design and execution rather than intelligence.
There was a big line of people standing just outside the small airport café, one of the very few eating establishments in this airport. Outside the entrance to a sat a sleek automated point-of-sale kiosk. It looked modern, efficient, and clearly intended to streamline ordering. Unfortunately, that was what people saw and naturally, a long line formed, all waiting to enter their orders. All this was made worse as there were two people trying to enter their order, and they were having problems (I listened in later and it turned out to be a language issue as English was not their native language, which raises a UX question but that’s for another time).
When Technology Creates the Problem
But here’s the twist.
I look past the line into the restaurant to the counter and see two employees working at the front distributing food but also see two POS terminals.
Inside the café? No line. Not a single person waiting.
Two staff members behind the counter, ready and willing to take orders the old-fashioned way. Walk in, order verbally, pay and grab your food, and go. The entire process took about 90 seconds and that included getting my food. Before I sit down with my food in hand, I look back, and the same two people are *still* trying to order and the line was getting longer. What was supposed to be a “fast” system was a painfully slow one. So I try to help a few people in line and announce:
“Hey, you don’t need to wait,” I said. “You can just go inside and they will take your order directly. No line”. One individual broke from the line and thanked me profusely, but the remainder stared at me as if I had just landed from another planet.
I eat a leisurely meal, finish, clean up, and leave before the people in line have even ordered.
Why?
- It wasn’t lack of evidence. They could see the outcome.
- It wasn’t lack of information. I had told them directly.
- It wasn’t lack of opportunity. The path was wide open.
- What they lacked was trust.
I was a random stranger offering new information that conflicted with what their environment seemed to be telling them. The kiosk was outside. The line was there. Therefore, the line must be the right thing to do.
Humans, it turns out, are excellent at following signals and terrible at questioning them.
The Kiosk, the Sandwich, and the Future of Agentic AI
Which brings me to healthcare and the wave of Artificial Intelligence and especially Agentic AI that will no doubt be talked about extensively at the HIMSS show taking place in Las Vegas this week.
Agentic AI systems promise something far more powerful than today’s passive tools. Instead of simply answering questions, they can take initiative. They can monitor patients, coordinate care, flag risks, automate administrative tasks, navigate prior authorization, and even orchestrate complex workflows across systems.
In theory, this could transform healthcare.
Clinicians could spend less time fighting bureaucracy and more time practicing medicine. Patients could experience real medicine delivered by their clinician and receive better care, smoother care journeys, and fewer administrative headaches.
But the uncomfortable truth is that technology isn’t the main barrier. Human behavior is.
Smart AI, Dumb Lines
Just like the airport line, healthcare is full of embedded signals. Clinicians follow workflows because that’s how the system has always worked and in many cases how it is mandated to be used. Patients follow instructions because the structure around them suggests that’s the correct path.
When a new signal appears, even a better one we often see it ignored.
That’s exactly what will happen to Agentic AI if we deploy it poorly.
If an AI agent quietly suggests an alternative care pathway but the electronic health record demands clinicians to follow a different workflow, guess which one wins?
If an AI system flags a risk but the alert arrives in a sea of alert fatigue, guess what happens?
And most of all if an agent recommends a change but the user doesn’t trust the source, the recommendation is DOA.
Technology doesn’t shape behavior. *Design does.*
Agentic AI will only deliver real value if we build systems that deliberately shape the behaviors we want to see.
That means designing environments where the right action is the easiest action. The path of least resistance and like the airline industry, which delivers six sigma level safety, is focused on failing safely.
AI recommendations must be integrated into workflows, deliver insights with transparency to build trust, continuous feedback and learning, and delivering positive outcomes to reinforce confidence. It also means acknowledging something deeply uncomfortable for technologists: intelligence is not enough. You can build the smartest AI in the world, but if the interface points people toward the wrong kiosk, they’ll still stand in the wrong line.
Standing in the Wrong Line
Right now, several forces are holding back the real impact of Agentic AI in healthcare.
Fragmented data trapped inside electronic health records. Regulatory uncertainty around autonomous decision-making. Misaligned incentives that reward volume rather than outcomes. And perhaps most importantly, instead of designing for the current system, stop being encumbered by historical systems and services, and create the change with a new approach and solution
Perhaps the biggest obstacle may simply be that we still design technology around the assumption that humans will adapt to it.
They won’t.
If we want Agentic AI to transform healthcare, we have to design systems that shape behavior the way airport signage shapes passenger flow, clearly, intuitively, and irresistibly.
Because if we get the design wrong, we’ll end up with the healthcare equivalent of that airport café:
A powerful tool sitting right in front of everyone…
While the entire system stands patiently in the wrong line.
You Can Build Brilliant AI… But People Will Still Queue for the Kiosk
