Build a Voice AI for After-Hours Property Management

Most property managers initially rely on standard phone trees or Interactive Voice Response (IVR) systems from providers like RingCentral. This approach consistently fails because tenants facing emergencies, such as a burst pipe or an HVAC failure in winter, are not inclined to navigate a "press 1 for X" menu. They often hang up in frustration and call repeatedly, escalating the issue and leading to negative reviews centered on poor response times. These basic systems cannot reliably distinguish a genuine emergency from a simple request, often misrouting critical calls or failing to record necessary details.

Moving beyond basic IVR, some firms attempt to use off-the-shelf conversational AI platforms, often integrated with Twilio. While an improvement, these platforms frequently struggle with the stateful, multi-step logic essential for property management. A single after-hours call might require checking a complex, property-specific on-call schedule (which often resides in a Google Sheet or internal tool), verifying the tenant's unit and lease status via an API call to RealPage or AppFolio, and then texting the correct, pre-approved vendor for that specific issue. Each step consumes resources, and the variable billing models can lead to unexpected costs, sometimes exceeding $400/month for a 300-unit building with frequent after-hours activity.

These generic tools are not engineered for business-critical workflows common in property management. When an API call to a system like Yardi fails, or a vendor is unresponsive, the workflow often simply stops, leaving the tenant in limbo and the property potentially at risk. There is rarely built-in logic to automatically retry a connection, escalate to a human property manager, or dynamically adjust routing based on real-time availability. This lack of resilience is a critical failure mode when dealing with property emergencies at 2 AM, directly contributing to the "siloed systems that do not communicate" pain point that plagues many property management operations.

Syntora's approach would begin with a detailed discovery phase to map your exact after-hours protocol into a robust state machine. This involves defining what constitutes a P1 emergency (e.g., fire, flood, active water leak into another unit), a P2 urgent request (e.g., no heat during cold months, locked out, running toilet), or a P3 routine ticket (e.g., dripping faucet, flickering light, amenity booking inquiry). This precise logic would form the core instruction set for the Claude API, guiding the conversational flow and ensuring accurate classification based on tenant input.

We would provision a dedicated phone number via Twilio and connect it to a FastAPI application hosted on AWS Lambda. When an incoming call is received, Twilio would transcribe the user's speech in real time and transmit it to our API. The FastAPI service would then send this text to the Claude API, which is highly capable of parsing natural language and returning structured JSON to classify the issue and inform the next conversational step. Syntora has extensive experience building document processing pipelines using Claude API for sensitive financial documents, and the same pattern applies to handling the nuanced inquiries of property management tenants.

If Claude identifies an emergency, the FastAPI application would use httpx to perform an immediate lookup. This lookup would query a client-provided data source, such as a Supabase database, an internal vendor management system, or a Google Sheet containing current on-call schedules, to identify the correct technician or emergency contact for that specific property and time. The system would then send a high-priority SMS alert via the Twilio API, including the unit number, tenant name, and a summarized description of the issue. The system would also integrate with client systems like RealPage or Yardi to verify tenant details before dispatch.

For non-emergency requests, the system would gather the necessary details from the caller and make a direct API call to your existing property management software, such as AppFolio, Buildium, RealPage, or Yardi, to create a maintenance ticket. Every call transcript, AI classification, and action taken by the system would be written to a log stream using structlog, providing a complete audit trail for all after-hours incidents and supporting financial reporting needs. The delivered system would be owned by the client, with full access to the codebase and infrastructure for ongoing maintenance and future enhancements.