An AI voice agent that handles after-hours maintenance calls, determines if there's a true emergency, sends out text alerts to the on-call team when needed, and logs every call automatically.
A property management company in Philadelphia manages around 500 properties. Tenants call about maintenance issues at all hours, and the nature of the call ranges from genuine emergencies (water main break, no heat in winter, gas smell) to non-urgent requests (closet door off the track, light bulb out).
The company had two options: pay for an after-hours answering service (expensive, inconsistent triage quality) or route calls to on-call staff (burnout, interrupted sleep, and the same inconsistent triage problem). In both cases, non-emergency calls regularly triggered unnecessary dispatches, while some real emergencies got delayed in the queue.
The core challenge was consistent, reliable triage at a volume and time of day when human performance naturally degrades.
We built an AI voice agent that serves as the after-hours maintenance line. When a tenant calls outside business hours, the voice agent answers and conducts a structured triage conversation.
The agent greets the caller and identifies the property and unit by address, unit number, or tenant name. This ensures the issue is tied to the correct property and the caller is a known tenant.
The agent asks a series of questions designed to determine if the issue qualifies as a true emergency. The decision tree covers categories like water (flooding, no water, water main break), fire and gas (smoke, gas smell, CO alarm), HVAC (no heat when below freezing, no A/C during heat advisory), security (break-in, lockout with safety concern), and electrical (sparking, outage affecting safety systems).
The agent uses natural language understanding to parse tenant descriptions, ask follow-up questions when needed, and categorize the issue. It doesn't just listen for keywords. It understands context ("there's water everywhere" vs. "the faucet drips a little").
Every call is logged to the company's Google Sheets with full notes from the conversation, including the property, unit, tenant, issue description, and triage classification. Confirmed emergencies trigger immediate SMS notifications to the on-call maintenance team. Non-emergency issues are logged for the next business day, and the tenant is informed that their request has been recorded and will be addressed during normal hours.
The voice agent provides consistent, reliable triage every night across all 500 properties. No fatigue, no missed calls, no inconsistent classification. Emergency calls are correctly identified and the on-call team is notified via text within minutes. Non-emergency calls are captured and queued without disturbing staff.
Unnecessary after-hours dispatches have dropped significantly. On-call staff only get woken up for real emergencies. And every call is fully documented with the conversation summary, making follow-up the next morning seamless.