Start with where time actually goes
Ask a facility or community manager where their week goes and the answer is rarely 'engineering'. It is finding documents, reading tickets, chasing updates and writing reports. These are the tasks where AI can help today.
Five realistic uses
Each of these supports a person rather than replacing their judgement:
- Search: asking for a warranty, certificate or last service report in plain language
- Summaries: a readable picture of open work, overdue PPM and escalating complaints
- Pattern detection: noticing the same fault recurring on one asset or floor
- Triage: suggesting a category and priority for incoming complaints
- Drafting: first versions of board updates and monthly reports
Why data structure comes first
AI can only be as reliable as the records underneath it. If assets are not named consistently, work orders are not linked to assets and documents sit in personal inboxes, AI output will be vague or wrong.
This is why a structured operating record — assets, work, complaints, contracts and documents in one place — is the real starting point. For older buildings this can be built through surveys and manual onboarding; newer buildings can add BMS and IoT data through APIs.
Keep people accountable
Maintenance, safety and compliance decisions carry legal and contractual responsibility. Good AI tools show their sources and leave the decision to a named person. Be wary of any product that suggests otherwise.
A sensible adoption path
Begin by consolidating records, then introduce search and summaries, then pattern detection. Measure success by hours returned to the team and problems noticed earlier — not by headline claims.