The practical answer
AI can help classify complaint descriptions and summarise repeated issues in a facility register. Reliable analysis requires consistent asset and location identifiers, agreed categories, and human review of proposed duplicates. A pattern should trigger investigation; it does not prove a root cause.
Decide which management question you need to answer
A complaint register can answer several different questions: which services generate repeated requests, where old tickets remain unresolved, or whether a repaired asset is being reported again. Choose one question before combining them into a dashboard.
For a monthly client review, a useful starting question is: “Which issues returned after a recorded resolution, and what evidence do we need to establish why?” This connects the analysis to an operational decision rather than a long list of keywords.
See the broader AI skills roadmap for facility managers for the relationship between data preparation, analysis and management review.
Clean the fields that determine what counts as a repeat
Keep the original complaint text and add working fields rather than overwriting it. At minimum, retain ticket ID, received date, site, floor or zone, service, asset ID if known, priority, status, closure time and a resolution note.
Normalise obvious spelling variations in site names. Separate “date reported” from “date entered”. Mark an unknown asset as unknown; guessing an asset can create a false recurrence.
| Case | Interpretation to review |
|---|---|
| Two people report the same open issue | Possible duplicate reports of one incident |
| The same asset is reported after a repair | Possible repeat failure |
| Similar wording at different locations | Similar category, potentially unrelated incidents |
| A ticket is reopened | A status change that must remain in the history |
Ask AI for classifications with reasons
Give the assistant a category list and an “uncertain” option. Request a proposed category, the supporting phrase from the complaint, and a reason for any proposed duplicate grouping. Review uncertain or high-impact cases manually.
Classify each complaint using the supplied service categories. Preserve ticket IDs and original descriptions. Propose possible duplicates only when site, location, asset and timing support the grouping. Explain each proposal and mark uncertainty. Do not merge, delete or close records. Separate repeated complaints from confirmed repeat failures.
Siemens documents duplicate work-order detection in its maintenance software. Applying a general assistant to a spreadsheet is a separate workflow: it still needs definitions, evaluation and review.
Work through a small example before using a full register
Ticket A reports a warm meeting room on Monday. Ticket B describes the same room while A remains open. Ticket C reports warmth in that room after A was closed on Wednesday.
A and B may concern the same incident. C may indicate recurrence, an incomplete resolution, changed occupancy, or another cause. Neither similar wording nor the closure date establishes the engineering explanation.
Review a sample yourself and compare it with the assistant's classifications. Record both missed matches and incorrect matches. A tool that aggressively groups everything may make the register look cleaner while hiding separate service failures.
Turn the findings into an action review
Prepare a short table containing the issue group, linked ticket IDs, evidence, unresolved questions, responsible reviewer and next review date. Present recurrence separately from total complaint volume so a busy location is not automatically portrayed as a poorly managed one.
Check whether changes in occupancy, reporting channels or service scope affected the counts. A new reporting app can increase recorded complaints without proving that service quality deteriorated.
Use the reviewed findings in a monthly facility management report. Where repeated issues concern equipment, connect the review to maintenance history and work-order evidence.
Practical questions
Can AI identify the cause of repeated complaints?
It can propose investigation questions. Root-cause conclusions need supporting operational or technical evidence.
Should duplicate tickets be deleted?
Preserve the audit trail. A reviewed grouping can link records to one incident without removing the original reports.
Source notes
Product capabilities were checked against the following primary sources on 3 November 2026. The workflows and worked examples are illustrative management exercises, not reported client results.
- Siemens: AI for maintenance teams — duplicate work-order detection and maintenance assistance.