A maintenance inbox combines descriptions, photographs, forwarded documents and frustrated follow-ups. An organised case brief can help the property manager understand what was reported without treating every message as a verified diagnosis. AI can assist with that preparation when the record keeps the tenant's wording, the property identity and the unresolved questions visible.
This proposed intake process does not decide liability, repair obligations or entitlement to reimbursement. It also does not diagnose hazards from a photograph. The property manager and appropriate specialists remain responsible for urgency, safe instructions, access and any decision about repair responsibility. The useful output is an attributable report that supports those decisions.
Capture what was reported and where it occurred
Begin with the property reference, unit, reporter contact and receipt time. Record the tenant's description without removing uncertainty. Water seems to be coming from upstairs is different from upstairs plumbing caused the leak. The first is an observation and a suspected source; the second asserts causation.
Ask for the location within the property, when the issue was first noticed, whether it is ongoing and what evidence the reporter already has. Request only information relevant to the manager's process. Do not ask someone to approach a suspected hazard or perform a test just because a model thinks the detail would make classification easier.
Preserve the original message and attachments with stable references. If an image has no reliable timestamp, label that limitation. A forwarded photograph may show an earlier incident or another unit. A confident description of an image should never replace a checked property match or the reporter's actual statement.
Separate categories that answer different questions
Create fields for reported issue type, location, possible urgency indicator and responsibility question. Plumbing, electrical and building fabric can be routing categories, but they do not assign liability. Likewise, an urgency flag is a request for prompt human assessment, not a definitive finding about severity.
Use the manager's approved escalation rules for reports mentioning immediate danger. Make the established emergency contact route available through the organisation's reviewed process, and do not delay that route to complete an AI form. This article deliberately does not prescribe technical safety instructions or a universal emergency threshold.
Source: OpenAI Structured Outputs describes schema-constrained responses. A proposed intake schema can keep observations, suspected causes and unresolved decisions in separate fields. Correctly shaped categories do not establish whether a fault exists or which person is responsible. Permit unknown and needs manager review results.
Preserve documentary evidence without inventing its meaning
Attachments may include a repair invoice, an inspection report or a handwritten description. Source: Azure Document Intelligence overview describes read and layout extraction as well as prebuilt document models. These capabilities can prepare text and structure for review. They do not determine whether a maintenance claim is valid or whether an invoice should be paid.
Keep the extracted text linked to the original document and page. Where a scanned amount or date is unreadable, record it as unreadable rather than supplying a likely value. A repair invoice attached to a tenant message can be evidence of a claimed expense; it is not automatic approval of reimbursement or proof that the work occurred as described.
Photograph interpretation needs its own evaluated method if the agency chooses to use it. The document extraction source does not establish a general property-damage diagnosis capability. The proposed first version can simply retain photographs and let the manager inspect them alongside the message.
Reusable maintenance case brief
Complete this proposed brief for the manager. Keep the original evidence accessible to authorised staff and avoid copying sensitive material into a broadly visible task title.
| Field | What to record | What it does not establish |
|---|---|---|
| Case identity | Case ID, property ID, unit and linked earlier cases | Identity is not inferred from a similar address |
| Reporter | Supplied contact and stated relationship | Authority or tenancy status is not automatically verified |
| Report | Original wording and concise attributable summary | Summary is not a diagnosis |
| Issue location | Room or area as reported | Unknown location stays unknown |
| Timing | First noticed, ongoing or intermittent, evidence time | Receipt time is not onset time |
| Evidence | Message, image and document references | Attachment presence is not factual validation |
| Uncertainty | Suspected cause, unreadable detail or conflicting report | AI cannot resolve by choosing the most plausible account |
| Urgency route | Triggered approved rule and person notified | Flag does not decide hazard severity |
| Access | Coordinator and stated restrictions | Reporting a fault does not grant entry permission |
| Responsibility question | What the reporter asks the manager to decide | No automatic liability or payment conclusion |
| Manager action | Assess, ask for detail, arrange approved inspection or close | Contractor work and reimbursement remain separate decisions |
A reusable manager handover is:
“Case [ID] concerns [reported issue] at [verified property/unit]. The reporter states [attributable observation]. Evidence includes [references]. [Specific uncertainty] remains unresolved. The approved urgency route is [recorded action or not triggered]. Please decide the next assessment step and resolve [responsibility/access question]. No contractor instruction or reimbursement approval has been issued by this intake.”
The completion check is that the manager can identify the property, reported issue, evidence, unanswered questions and next decision without reconstructing the entire conversation. If the property match fails, the brief remains incomplete instead of being sent to a guessed portfolio owner.
Compare an ordinary report with conflicting evidence
In a hypothetical normal case, a tenant at unit M-12 reports a dripping kitchen tap noticed that morning and supplies one photograph. The workflow creates a case, records the observation and links the image. Under the manager's approved rules, it routes the report to the ordinary maintenance queue. The manager assesses what inspection or instruction is appropriate; the intake does not state that the tenant or landlord must pay.
A second hypothetical message says the ceiling is leaking again but includes a photograph without a unit reference. The sender has previously contacted the agency about two properties. The workflow asks which unit is affected and preserves the reported urgency for human assessment. It does not attach the image to whichever property has a previous water-damage case.
Now imagine the same report reaches the inbox by email and a messaging channel. Stable message or source identifiers can reveal repeat processing. Similar descriptions alone might concern two separate faults. Link suspected duplicates for manager review rather than closing one automatically. When a later message says the leak stopped, add that update to the case; do not erase the earlier report or conclude the cause has been repaired.
Route actions through a manager and evaluate the intake
Source: n8n human review documents approval or denial before selected agent tools execute. A workflow may use that boundary for a contractor message or record update. Approval tooling does not supply liability rules or prove that an inspection is safe. The manager's process and the implemented permissions must cover those decisions.
Evaluate case preparation against manager-labelled examples. Include wrong-unit photographs, missing onset times, conflicting descriptions, duplicate messages and unavailable attachments. Check whether summaries retain uncertainty, whether urgency reports reach the appropriate human path and whether any draft accidentally instructs work or assigns responsibility. Compare the manager's effort and corrections with the current intake process before making a benefit claim.
Questions about maintenance classification
Can a category such as tenant damage determine payment responsibility?
No. It embeds a conclusion that this intake is not equipped to make. Prefer a neutral issue description and a separate responsibility question for the manager. Evidence, the relevant agreement and applicable professional judgement may all be needed before a decision. Do not let a routing label become an accounting or legal outcome.
Should the assistant ask the tenant to inspect a dangerous fault?
It should follow the manager's reviewed contact and escalation process rather than improvise technical instructions. Missing information can remain unresolved when collecting it would be inappropriate. The case should retain the reported concern and the action taken through the established urgent human route.
What if the report contains a repair bill already paid by the tenant?
Retain the original bill and the tenant's stated request. Document extraction may help prepare its details for review, but the bill does not automatically establish reimbursement entitlement or approval. The manager and finance owner decide the next step through their approved process, with any unclear dates or amounts still visible.
If your business needs help defining this process, explore AI agents for estate agencies, the wider AI automation services, and our custom-agent workflow guide. The agents and automation comparison and custom AI agent glossary explain the terms. To discuss your records and approval rules, get in touch.