The best AI in healthcare removes friction quietly. The worst demos beautifully and deploys terribly. Here’s how I tell them apart.
Every few years, healthcare technology falls in love with a new word. I’ve been through enough of these cycles to recognize the pattern: the word arrives, the demos get louder, the budgets get bigger, and somewhere in the noise the actual problem we were trying to solve gets a little lost. Right now the word is AI. And most of what I see marketed as AI in healthcare is theater.
I don’t say that as a skeptic. I say it as someone who has spent a career betting early on the technologies that turned out to matter — wireless and IoT before they were obvious, Salesforce as a platform when building a healthcare company on it still raised eyebrows, and now AI. I believe in it. Which is exactly why the hype bothers me.
Here’s the pattern I’ve learned to distrust: AI that demos beautifully and deploys terribly. The patient-facing chatbot that dazzles in a keynote and frustrates a real person at 11pm. The predictive model that’s technically impressive and operationally useless because no one designed the workflow around what it predicts. The gap between “look what it can do” and “look what it actually changed” is enormous, and most buyers don’t find out which side they’re on until after the contract is signed.
The AI that actually earns its place tends to be the opposite of flashy. It’s the automation that quietly removes a step a human used to do by hand. It’s the layer of software that gives a person back twenty minutes so they can spend it on the part of the job a machine can’t do. You don’t demo that. You barely notice it. And that’s the point — the best AI in healthcare is the AI you don’t notice, because it’s busy removing friction instead of drawing attention to itself.
So when a leader asks how to evaluate an AI investment, I’ve stopped starting with capability. I start with friction: what specific, repeated, human-hours-consuming friction does this remove? If the answer is crisp, the technology usually earns its keep. If the answer is a vague gesture at “insights,” I get cautious — because AI that can’t name the friction it removes tends to add one instead.
I’m genuinely optimistic about what AI will do in healthcare over the next decade. But I’ve watched too many hype cycles to confuse motion with progress. The organizations that win won’t have the loudest demos. They’ll be the ones honest about what problem they were solving, disciplined about keeping humans where humans belong, and patient enough to build the boring, reliable tools that never make the keynote but quietly change how the work gets done. That’s the AI I’m interested in — the kind you don’t notice, until you try to go back to work without it.
Tammy Hawes is a healthcare technology executive, founder, and operator based in Brentwood, Tennessee. She founded Virsys12 in 2011 and led its 2025 acquisition by HealthStream, where she now serves as VP, Payer Solutions. She writes on AI in healthcare, leadership, and the founder’s path. More about Tammy →
Leave a Reply