Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

What agentic AI could fix for healthcare customer retention

The potential to scale customer engagement with personalization and precision.

[Photo: Getty Images]

Profile photo of Richard Mackey

Chief Technology Officer · CCS

Consider a hypothetical patient we’ll call Marie. She’s 65 and hasn’t reordered diabetes-management supplies in three weeks. She’s missed her care coach’s phone calls because she’s working two jobs with odd hours. And regular reordering feels like one more task she doesn’t have time to manage, especially since she’s wondering whether she can afford to keep up with her medical supplies at all.

Your customer retention model has already flagged Marie at high risk of abandoning her treatment plan. But the model doesn’t answer what happens next. What if the patient doesn’t respond to the next reminder email? Or the one after that? Chances are, the system will stop trying to engage at some point, turning an accurate signal into a failure to drive behavior change. 

I lead technology at a company that distributes diabetes supplies, and I see this as the central new challenge for healthcare-related AI adoption: customer engagement and retention.

Companies in every industry face a similar challenge: They believe seeing customer retention issues is the same as managing them. However, without effective strategies, the underlying problem will keep happening at the same rate, time and again.

A false sense of control

Most modern organizations have spent the past several years developing truly impressive customer retention models trained on behavioral signals, real-time data, and increasingly fine-tuned customer attributes. They can identify an at-risk customer days or weeks before the breaking point.

But actions after the flag goes up haven’t kept pace. Predictive output flows into static systems, including reminder sequences, batch campaigns, and persona-based outreach designed for populations, not individuals. A model that knows exactly why a specific customer is about to leave might trigger a message that could have been sent to anyone.

This leaves a gap between prediction and personalized, real-time response. And that’s where customers then disappear.

What behavior change actually looks like with agentic AI

Let’s return to Marie. But this time, she’s receiving her diabetes-management services from a company using an agent-driven model to monitor and manage her condition, where the same risk factors trigger a different sequence.

None of this exists yet, but here’s how it would work if it did.

The agent-driven model would recognize the lapse pattern and initiate a tailored interaction, with Marie’s consent. For example, if Marie doesn’t reorder her continuous glucose monitor supplies in a timely manner, an AI agent would reach out (as appropriate and within regulatory guidelines) when she’s home from work at night. When she engages, the agent would walk her through reordering in a few detailed steps.

When hesitation surfaces, the agent would shift to brief, targeted messaging about why consistency matters to her health, delivered with empathy. If Marie raises financial concerns, the agent would pivot to connect her with supportive resources and need-based options for discounted rates. The agent would base each real-time response on her reaction.

If Marie disengages, the system may adjust its timing, tone, or outreach channel based on her engagement patterns, offering a behavior-shifting, responsive coaching conversation the moment her actions start drifting.

Every industry that has built a retention model has the data and predictions. The ability to act on it individually, in real time, and at scale? That’s still missing for most companies, as no one has yet built AI agents that can offer adaptable, personalized interactions.

The constraint was always time, not intelligence

Human teams can craft good interventions and train agents to engage customers in a responsive, empathetic, and informed manner. But human-only teams experience time and capacity constraints, especially in industries where yearly enrollment deadlines or other seasonal events can cause a predictable but unmanageable logjam of traffic.

Someday, when AI agents augment the human workforce, they could handle far more of these tasks and expand capacity.

AI agents could reinvent the outreach equation by engaging every flagged individual with consistently high-quality interactions, no matter when or where the patient is ready to engage.

Every exchange generates data that forms patterns to guide the next interaction. When a situation requires human judgment, or a customer requests to speak to a human for any reason, the system would escalate so specialists handle the cases that genuinely need them.

Companies will see real retention differences only when redesigning their engagement models around continuous, individualized interactions rather than episodic outreach.

People will still generate the protocols and pathways and manage them when necessary. What changes is the ability to execute those protocols for every at-risk customer in an equitable and accessible manner, not just for the ones who happen to get a call back within traditional business hours.

Marie needs someone (or something) that understands her specific situation, walks her through the next steps at her own pace, and stays with her through next month’s cycle, and the one after that, and the one after that.

Prediction models might get Marie to the treatment doorway, but AI-powered agents may well greet her and help her step inside.

Richard Mackey is chief technology officer at CCS Medical.