Guest Post by Audrey Campbell, MPH, founder of The Audrey Aesthetic
Enterprise AI is having its reckoning. After two years of pilots and record spending, the returns are hard to find. MIT researchers who studied hundreds of enterprise deployments concluded that 95 percent of generative-AI pilots had produced no measurable profit-and-loss impact. Deloitte surveyed more than 1,800 executives and found only 6 percent could point to a use case that paid for itself inside a year; most put satisfactory returns two to four years out. PwC’s global CEO survey landed in the same place. A majority of chief executives reported neither higher revenue nor lower costs from AI over the past year, and only 12 percent could claim both.
The reflex is to blame the technology. It isn’t the technology. Over the past two years I have evaluated more than sixty AI-native and AI-enabled tools built for healthcare, and the ones that work are not meaningfully smarter than the ones that don’t. What separates a deployment that pays off from one that stalls is almost never the model. It is the organization around it.
I know this because I see it happen in a part of healthcare the AI conversation tends to overlook: the medical aesthetics practice. A cash-pay, consumer-facing, owner-run business, often with no data-science team and no innovation office, is pulling measurable value out of AI while far larger and far better-resourced companies can’t.
Start with why the setting matters. U.S. medical aesthetics is now a market north of seventeen billion dollars, growing by more than a billion a year, and the American Med Spa Association describes it plainly: largely cash-pay and recurring, with loyal patients and predictable margins. There is no insurer to hide behind. Every dollar comes from a consumer who chose to spend it, and the owner feels the result of a good or bad decision within the month. It forces the kind of clarity that large enterprises spend fortunes trying to manufacture.
Here is what the practice does that the enterprise can’t seem to.
One person owns the decision. In a practice, the owner evaluates the tool, approves it, and lives with the result. There is no steering committee, no cross-functional working group, no pilot that outlives the sponsor who championed it. Enterprise AI dies in exactly that gap. The most consistent finding across the failure research is not technical, it is organizational: undefined ownership, and no one accountable for the outcome. A practice never has that problem, because one person is accountable and it is the person who signs the checks.
The workflow gets rebuilt around the tool, not the other way around. When a practice adopts an AI system for patient follow-up or intake, it changes how the front desk actually works and retires the old process. Enterprises do the opposite. They bolt AI onto a legacy workflow, leave the workflow intact, and wonder why nothing moves. Survey after survey names the same obstacles: data that isn’t ready, integration with existing systems, and processes that were never redesigned for the tool to do its job. A small operator redesigns by instinct because she can see the whole workflow at once. A large one often can’t see it at all.
And there is always a number. The owner isn’t tracking “AI adoption.” She is watching rebooking rate, or no-show rate, or how fast an after-hours inquiry gets a real reply, and she can tell you within a day whether the tool earned its place. The enterprise measurement problem is well documented: leaders can’t prove a pilot worked because the data layer to prove it was never built. The practice has one owner, one metric, and an honest answer.
None of this makes the practice a model to copy wholesale. The same speed that lets a practice move fast can also let it skip a safeguard. I have seen practices on the verge of running protected health information through a consumer AI account, or handing a patient-facing bot the keys with no path to escalate when someone types something urgent. That is the shadow side of ownership without governance, and at scale it stops being hypothetical. Gartner forecasts that by 2027, forty percent of enterprises will demote or decommission autonomous AI agents after discovering governance gaps once those agents reach production. The small operator’s instinct for speed is exactly what enterprise governance exists to restrain.
So the lesson isn’t “be more like a med spa.” The practice is right about ownership, workflow, and measurement, and lucky about governance. The enterprise is the reverse. The organizations that actually capture AI’s value will be the ones that hold both at once: the owner’s clarity about who decides and what number matters, paired with the discipline about data, privacy, and escalation that a regulated business demands.
The deeper reason the practice keeps winning is that it treats AI as infrastructure. A system that either moves a number the owner cares about or gets switched off. That is where the entire enterprise conversation is finally arriving, years and billions later. The companies that get there first won’t be the ones with the biggest models or the largest AI budgets. They will be the ones that learn to decide the way an owner does.
Audrey Campbell, MPH, is the founder of The Audrey Aesthetic, an advisory on AI implementation in medical aesthetics. She has evaluated more than sixty AI-native and AI-enabled tools built for healthcare and writes on where AI creates real operational value in patient-facing practices.
