Every conference you go to this year will tell you to focus on retention. Every webinar, every breakout, every keynote. And they’re right that it matters. Keeping the patients you already have is far more profitable than winning new ones, and most owners understand that by now. What nobody says out loud is that retention, the way it runs in almost every practice today, is manual work with a technology label on it.
So here is the distinction I keep drawing for people, because most get it wrong.
If you pull a list of everyone who didn’t rebook and drop them into a retargeting campaign, that is not AI. If you build a segment of patients who are 90 days overdue and trigger an automated message, that is not AI either. Those are rules. You wrote the rule, the system follows the rule, and the rule never changes until you change it. Useful. Worth setting up. But it is automation, and automation is not what the industry means when it says the word AI.
When I explain this, the response I get most often is some version of “oh, we already do that, we have a target segment set up.” That is the exact moment I know it hasn’t landed.
The one thing to ask: does it learn
The simplest way to tell automation from AI is to ask whether the system gets better on its own.
A target segment doesn’t. It sends what you told it to send, when you told it to send it, and it is exactly as smart a year from now as it is today. Real AI watches what happens after it acts. It learns which messages get opened and which get ignored. It learns the times a specific patient actually responds. It adjusts the next message based on the last one, using information you never handed it, because it gathered that information itself.
That is the whole difference. Learning versus following instructions. Everything else is branding.
What agentic actually means in a practice
The word doing the real work here is agentic. An agent has agency. It doesn’t wait for you to build a list and hit go. It sees that you have low utilization coming up on Thursday and reaches out to the patients who could fill it.
And it reaches out with context. A real agent sees that a patient has come in for injectables four times and has never booked a facial or tried a single spa service, so it reaches out to that patient about the spa. It isn’t sending a generic “we miss you” to everyone 90 days out. It reads the treatment history, finds the gap, and goes after the specific patient most likely to say yes to the specific thing.
That is what moves your numbers. The three levers that actually drive retention are membership penetration, treatment mix, and rebooking rate. An agent reading individual history and acting on treatment gaps works all three at once. A rule-based segment can’t do that.
Where the human stays
I want to be careful here, because agentic can start to sound like you fire your front desk and let the software run the practice. That is not it.
The best use of an agent is to protect your team’s time for the patients who need a person. The agent handles the high-volume, repeatable outreach, then routes the high-value patient, the one worth a personal call, to your front desk associate to handle directly. The AI decides who should get a human and makes sure they do. The person doesn’t leave the loop. The person gets pointed at the patients who matter most.
What to do with this
Almost everything being sold to you as AI retention right now is a target segment with a better logo. The real thing is rare, and it is worth evaluating carefully before you pay for it. The test holds up every time: ask the vendor how the system learns and gets better. If the answer is a list you build and a message it sends, you already have that, and it isn’t AI.
If you want to see what it looks like when it’s built right, that’s what I write about on the Substack.
Warmly,
Audrey
Subscribe now
Originally published on Substack.