By CRAIG HAUBEN
Ask anyone outside of healthcare who is resistant to clinical AI and you’ll get a sure answer. The older doctors. Those who spent thirty years gaining experience and now see a machine that is coming for it. The story writes itself, which should have been the first clue that I was wrong.
I’ve spent thirty years in healthcare and now run a company that builds and runs AI within provider and payer organizations. At Clutch we use AI data analytics to solve engagement challenges. WHO is the patient today? What message will they receive? When do you want to read it? If you do it right, you can drive the kind of sustained behavior change that drives clinical outcomes like medication adherence, care plan adherence, and closing gaps.
So I’m not working from theory. I look at this land in real workflows, and this is what I see. The doctors most excited about AI tend to be those who have been doing the job the longest. The resistance comes from elsewhere. If you manage a health system, that difference should change how you plan your next deployment.
Start with the adoption numbers, because they already reveal the history of resistance. The latest AMA survey found Four out of five doctors now use AI in practice.up from 38 percent in 2023. That is not a profession that opposes a threat. That’s a profession that found something useful.
Now the veterans. A doctor with three decades in a specialty can see, better than anyone, what these systems are good for. Pattern recognition at scale. Catching something that should have been pointed out two visits ago. Bringing to light what was already reflected in the data: the missing finding on last year’s images, the lab trend over eighteen months that seemed normal, one value at a time, the three emergency department visits in six weeks that no one had time to connect.
This is not hypothetical. He Studying the nature of Google’s breast cancer detection system showed a 9.4 percent drop in false negatives for American patients, the cancers that human readers missed. He biggest NHS assessment to dateamong 175,000 women, found that AI detected more invasive cancers with fewer false positives than human readers. The damage that these systems pursue, information that existed and was never connected, is something that experienced doctors know only too well. They have spent their careers watching how their absence hurt people.
Here’s one of our own work. We’re working with a national government program payer on some of their hardest-to-engage members, the high-intensity ones who need contact four or five times a day for six months or more. We achieved 95 percent engagement, as measured by the customer, and 93 percent adherence. The result was an average 0.8 drop in HbA1c and an 18 percent reduction in symptoms.
When a system eliminates the mechanical load so that judgment work receives more attention, the thirty-year-old clinician does not feel threatened. They feel relieved. Their expertise is trial, not data recovery, and they have always known the difference.
Now look where fear really lives. It comes from the middle.
The people who built their careers on being the synthesizer, the translator between systems, the one who extracted information from six places and assembled it into an image. That role is under real pressure, not clinical judgment. Anthropic’s labor market research points the same way, finding that exposure to AI is concentrated in exactly this type of assembly work rather than in high-judgment roles. The synthesizer is afraid. And the synthesizer is right, because synthesis is what these systems do best.
Both answers are rational. That’s the point. Its workforce is not divided between the enlightened and the fearful. It’s divided by what people do all day, and the line doesn’t go where conventional wisdom says.
If you run a health system, that has three consequences.
First, your implementation advocates are not who your consultants think they are. The standard manual recruits young doctors as AI ambassadors, based on the theory that digital natives adapt faster. Instead, recruit thirty-year veterans. They have credibility, they can tell exactly where the system helps and where it can’t be trusted, and their word carries different weight in the staff room. One skeptical experienced doctor turned into a precise and conditional advocate is worth more than ten enthusiastic residents.
Second, people in the synthesis layer deserve honesty, not slogans. Telling a care coordinator or utilization review nurse that AI will simply make their job easier is a way of destroying your own credibility, because they can see the mechanism as clearly as possible. The honest conversation revolves around what parts of the role are moving to the machine, what the role becomes after that, and what the institution will do to help people bridge that gap. Most organizations are not having that conversation. Those who do will stay with their best people. Those who don’t will lose them at exactly the wrong time.
Third, stop measuring adoption and start measuring trust, in both directions. Nearly all Fortune 500 companies now track employee AI usageand healthcare is copying the habit. Usage is the wrong metric. A system that doctors use reluctantly by mandate is a risk. A system that doctors trust beyond its actual performance is bigger. He AMA Sentiment Data Captures correct posture better than any board. About two in five doctors say they are equally excited and concerned, and that ambivalence is not a problem that needs to be fixed. It is the right response to a powerful tool with uneven performance, and it is exactly the provision on which good governance should be built.
Veterans are your asset here too. The doctors most excited about these tools are often the most precise about their limits, because true expertise includes knowing what the tool can’t do. Build your monitoring around that accuracy rather than utilization dashboards, and you’ll have an early warning system staffed with the people best qualified to run it.
Coverage of AI in medicine continues to offer two stories. The machine that replaces doctors, or the machine that destroys medications. The people who run things don’t live in either. The real version is more granular, in places really good, and starts by noticing that the people we expected to resist are the ones quietly showing us how to use it well.
That gap, between the two clean stories and what happens on the ground, is what prompted me to write the book. AI: migration is a novel about how AI is revolutionizing work. Every AI system and clinical event it contains is drawn from the documented record, so while the characters are fictional, the AI stories are real.
Craig Hauben is CEO of Clutch and has spent thirty years as a healthcare operator and executive. his novel AI: migration publications in July 2026.
