The Challenge

A clinic was working with a big group of patients who had pre-diabetes and type 2 diabetes. Each patient had $100/month in insurance coverage for care, and the clinic could reach them by phone or text.

The care team kept reaching out, encouraging people to eat better, move more, and keep up with check-ups. But there was a big problem: many patients weren’t booking their first visit.

The team was doing the right things, but as the patient list grew, they needed a way to scale outreach and actually turn it into action.

The Solution

The clinic brought in Puppeteer’s AI assistant to help.

This wasn’t just another reminder system. The AI could text and call patients, remember past conversations, and adapt its tone and language. It used small nudges from behavioral psychology to guide patients toward one goal: book the visit.

Outreach became scalable, consistent, and more personal—turning follow-ups into real appointments.

The Process

Target: 30 bookings per day.

Setup

We began by building the first version of the agent with a script tailored for this clinic. To ensure reliable communication, we purchased local phone numbers from the patients’ area codes and registered them properly so they wouldn’t be flagged as spam.

Learning from the Team

Next, we spoke with the staff member who had been making manual calls. Their experience revealed weak points in the existing process, and we used those insights to guide our first round of improvements.

Handling Edge Cases

One major issue was language. Many patients weren’t comfortable speaking in English, which blocked engagement. To solve this, we introduced real-time language switching so patients could communicate in the language they preferred.

Experimenting

We ran a series of experiments to refine outreach. We tested different call cadences, making up to three attempts per patient, and tried various times of day—morning, midday, and afternoon. Since results were similar, we settled on hourly call batches. We also experimented with different voices, finding that patients responded most naturally to a female voice. For messaging, we A/B tested different approaches and learned that building trust early, by presenting the AI as part of the health plan, was the most effective strategy.

Expanding

After refining the process with adult patients, we extended outreach to parents of pediatric patients. Engagement improved even further, showing that family-level outreach could drive stronger results.

The Results

1. Bookings jumped from 5 per day to 30 per day—a 6x increase.

2. The AI scaled beyond what the call center team could handle on their own.

3. Outreach became consistent, personalized, and more effective.

The Conclusion

By using Puppeteer’s AI assistant, the clinic turned outreach into measurable results. Iterations like language switching, trust-building, and smarter cadences made the model scalable.

What started as a pilot showed how AI can extend care teams’ reach and help more patients take the first step toward better health.