Implement an AI-Powered Patient Engagement and Recall System

Most dental offices use the recall module built into their PMS or a bolt-on service like Weave or Demandforce. These systems send generic, one-way SMS blasts: "Hi Jane, you are due for your cleaning. Please call us to book." Patients ignore these messages because they are impersonal and require them to stop what they are doing to make a phone call.

When patients do reply via text, the interaction breaks. A patient might respond, "I have new insurance, is it in-network?" or "Any chance for a Tuesday morning next month?" Standard recall software cannot answer these questions. It creates an alert for your front desk staff, forcing them to manually look up the patient, check the schedule, and text back. This defeats the purpose of automation, turning a time-saving tool into a disorganized ticketing system.

The core failure is that these tools are not conversational; they are notification systems. They cannot access the live schedule to offer specific times, cannot answer simple questions, and cannot understand intent. This leaves your staff to manage hundreds of disjointed text threads, while thousands of dollars in potential hygiene revenue sits unscheduled in your patient database.

Syntora would approach the development of an AI-driven patient recall system through a phased engagement, focusing on architectural clarity and verifiable outcomes.

The initial step would involve establishing a secure, read-only connection to your on-premise or cloud PMS database. Syntora would implement a Python script to run daily, syncing overdue patient records and extracting key details such as last visit date, scheduled family members, and previous appointment notes. This extracted data would populate a dedicated Supabase database, creating a clean, structured source for the AI.

For the conversational AI agent, Syntora would implement the Claude 3 Sonnet API. This API would draft personalized messages for each patient, referencing their history. For example, a message might read: "Hi John, it's been just over 6 months since your last cleaning with Dr. Smith. We have a few spots open with your hygienist, Sarah, next week." Syntora has built document processing pipelines using Claude API for financial documents, and the same pattern applies to extracting relevant patient history and generating appropriate responses for dental documents.

The core of the system would be a FastAPI application deployed on AWS Lambda, responsible for managing the two-way SMS conversation. When a patient replies, the AI agent would interpret their request, query the live schedule via the PMS connection, and offer 2-3 specific, available appointment slots. The system would allow patients to book, ask for different times, or ask questions, guiding them through the scheduling process.

Once a patient confirms a time, the system would write the appointment directly back into your PMS schedule. The infrastructure costs on AWS Lambda for an office with 10,000 patients typically run under $40 per month. Syntora would define clear escalation protocols, ensuring your front desk staff is involved only for complex queries that require human judgment.

A typical engagement for this complexity involves a build timeline of 8-12 weeks, including initial discovery, system design, development, and testing. For successful deployment, the client would need to provide access to their PMS, document existing scheduling policies, and collaborate on validating the AI's conversational flow. Deliverables would include the deployed AI system, comprehensive technical documentation, staff training, and an agreed-upon support and maintenance plan.