Where the work starts
Extract property reference, reported fault, access restrictions and urgency signals.
Start with a fictional example that resembles normal work, but remove identifiable customer, employee or business information.
Turn scattered maintenance messages into useful repair tickets while keeping urgency decisions with the team.
Extract property reference, reported fault, access restrictions and urgency signals.
Start with a fictional example that resembles normal work, but remove identifiable customer, employee or business information.
Draft a ticket summary and questions for the tenant.
The output is a proposed working document, not a verified record or a completed business action.
ACT AS: a careful internal drafting assistant supporting property management coordinators. GOAL: Extract property reference, reported fault, access restrictions and urgency signals. PRODUCE: Draft a ticket summary and questions for the tenant. INPUT: [replace with a fictional scenario containing a few deliberate missing details]. DO NOT: assume facts, make commitments, reveal personal information or act without approval. When uncertain, explain what must be checked. REVIEW RULE: Escalate safety risks immediately; never diagnose a hazard or promise a contractor visit. FORMAT: (1) request summary, (2) essential missing details, (3) draft for a staff member to review, (4) safety or uncertainty notes.
Never copy confidential records into a service your employer has not approved. Treat AI drafts as untrusted until someone checks them.
Test a leaking tap, an electrical smell and a report without an address; compare routing choices.
Repeat with one missing field and one deliberately contradictory detail. The system should highlight ambiguity rather than conceal it.
Escalate safety risks immediately; never diagnose a hazard or promise a contractor visit.
Record who has the authority to review the result, correct errors and decide whether the draft can be used.
Track time spent requesting missing repair information. using a small fictional or securely controlled sample. Count the human effort required to verify the result, the number of corrections and the cases where the draft fails. If the checks take longer than the original task, simplify the process instead of automating more.