Created an AI maintenance intake system that classifies repair urgency, gathers evidence, and dispatches vendors faster.
Challenge
A property manager handling hundreds of units received maintenance requests through phone, email, portal forms, and text. Staff manually classified urgency, asked tenants for photos, contacted vendors, and answered status calls. True emergencies could be buried inside routine requests.
Workflow approach
We built a maintenance triage assistant across voice, SMS, and web chat. It asks structured questions, collects images, classifies urgency, checks property and lease data, creates vendor-ready work orders, and sends tenant status updates. Emergency categories trigger immediate escalation to on-call staff.
Observed outcome
Triage time fell 61%, status calls dropped 48%, and vendors received better work orders with photos and access instructions. Emergency issues were surfaced faster while routine repairs moved through a consistent workflow.
Outcome metrics
- Faster triage
- 61%
- Fewer status calls
- 48%
- Vendor response speed
- 2.2x
- Requests auto-classified
- 89%
Systems involved
- GPT-4o
- Twilio
- n8n
- Propertyware API
- Airtable
- Slack