Automate Maintenance Triage and Vendor Dispatch with AI

Most small property managers start with a shared Gmail inbox and spreadsheets. This breaks down quickly. An urgent "no hot water" email gets buried under routine requests, leading to frantic calls and angry tenants. There is no central dashboard to see what is open, who is assigned, or when a job was completed.

Dedicated property management software like AppFolio or Buildium is a step up, but their built-in workflows are rigid. They can create a work order from a tenant portal submission, but the dispatch logic is basic. It cannot handle rules like, "For a plumbing emergency after 6 PM on a weekend, text our three on-call plumbers in sequence until one accepts." The property manager still has to manually make the calls for any non-standard request.

Connecting a web form to a Trello board using Zapier seems like a solution, but it only moves the problem. The Zap creates a card, but a human still has to read it, check a spreadsheet for a qualified vendor, call them, and then manually update the Trello card. Each request burns 3-5 Zapier tasks just to log the data, pushing your bill to over $100/month for a glorified notification system that does not perform the actual work.

Syntora's initial engagement would involve a discovery phase to understand your current data landscape and operational procedures. We would identify relevant data sources, which typically include existing property management APIs like AppFolio, email systems such as Gmail for maintenance requests, and vendor contact lists often managed in spreadsheets. This historical data and vendor information would then be structured and loaded into a database, for instance, a Supabase Postgres instance, forming the foundation for the AI model.

The core of the proposed system architecture would be a FastAPI service. This service would be designed to listen for new maintenance requests, either through webhooks from a property management platform or by polling a dedicated maintenance inbox. Within this service, the Claude API would be integrated to parse and interpret tenant messages. Similar to how we've built document processing pipelines using Claude API for financial documents, this approach can effectively classify issue types (e.g., plumbing, electrical, HVAC) and assess urgency based on the message content.

Upon classification, the dispatch logic would query the Supabase database to identify suitable vendors. For example, a high-urgency plumbing issue would trigger a search for vendors tagged with 'plumbing' and '24/7-emergency' within your defined vendor pool. The system would then use a communication API, such as Twilio, to notify a selection of matching vendors, allowing for an automated assignment based on their responses.

For deployment, the FastAPI application would typically be packaged as a Docker container. It could then be deployed on a serverless platform like AWS Lambda, allowing for scalable and cost-effective operation. This serverless architecture is well-suited for event-driven workflows like maintenance request processing. We would integrate structured logging, perhaps using structlog to AWS CloudWatch, and configure monitoring alarms to ensure the system operates reliably and to notify your team of any potential issues, such as API error rate thresholds or latency spikes. The delivered system would be a custom-engineered solution, designed to integrate with your specific environment and workflows.