Property maintenance automation: request to resolution

The problem and current manual process

Photos arrive in the agent's chat, the contractor gets a forwarded message and the owner is asked for approval separately. The tenant follows up because there is no shared job status.

Who should use this approach?

Property operations teams with agreed repair categories, a current property register, a contractor list and named people responsible for approvals.

When it is not a good fit

Do not launch automated dispatch before agreeing emergency escalation and spending authority. Where property ownership or tenancy records are unreliable, repair the register before letting a workflow disclose details or book access.

Proposed workflow

Systems involved: Property records, intake form or messaging, contractor calendar, accounting.

1. Create a maintenance record with property, unit, issue, photos and access instructions; link repeat reports to the existing job.
2. Route by the property's agreed rules, request spending approval where needed and book an available contractor.
3. Track attendance and evidence of completion, request tenant confirmation and match the contractor invoice to the approved work.

Synthetic example: two tenants report the same communal leak. The system links both reports to a candidate job, while a coordinator confirms they concern the same fault. A later quote above the approved budget returns the job to approval rather than silently booking extra work.

Where AI is useful

AI can draft a summary of the reported issue and suggest a category. It should preserve the tenant’s original wording and indicate missing information. It must not diagnose a safety issue, declare the property safe or decide that an ambiguous report is non-urgent.

Where deterministic rules are better

Use explicit rules for property access, job identifiers, spending thresholds, authorised approvers and contractor availability. A request is not closed simply because a contractor marked a visit complete.

States and human approval

1. Received: preserve the report and property reference.
2. Triage: a coordinator confirms category, urgency and access requirements.
3. Awaiting approval: attach the current scope, quote and authorised approver.
4. Scheduled: confirm contractor, tenant access and visit window.
5. Completion review: collect work evidence and outstanding issues.
6. Closed: record the responsible person’s decision and remaining follow-ups.

Urgent or ambiguous reports are escalated to the designated person. Emergency instructions remain visible; costly work requires owner approval.

Failure modes and recovery

Repeated report

Suggest a related job; do not merge unrelated repairs solely because the text is similar.

Quote changes after approval

Invalidate the old approval for the changed scope and request a new decision.

Contractor cancels

Reopen scheduling and notify the coordinator; keep the original tenant commitment visible.

Urgent or ambiguous report

Show the agreed emergency instructions and escalate to the designated person.

Security and privacy boundaries

Restrict unit details, tenant contact information, photos and access instructions to the people assigned to the job. Avoid putting door codes into general chat. Remove unnecessary personal details before any AI summarisation and agree retention for repair evidence separately from access instructions.

What a pilot should prove

• A changed quote cannot reuse an earlier approval.
• Two staff members cannot confirm conflicting appointments for one job.
• An unassigned urgent report remains visible to the designated coordinator.
• Closure records who reviewed the work and when.
• Tenants cannot see other units’ private records.

Intended operational effect

Tenants, agents and contractors have a clear next step, while property managers can see stalled jobs.

Measure a representative baseline, then compare the same scope during the pilot. Include review, exception handling and maintenance effort. Proposed measures: time to assign a contractor; open jobs beyond the agreed deadline; repeat follow-ups per repair.