Maintenance requests with an AI service bot: what works and what to watch

Maintenance is where tenants judge a property manager most directly. A service bot, a chat assistant that takes requests, can improve it when it is designed with care. Here is what it should do and where it needs limits.

What a good bot does

Collects complete information. Instead of "AC not working", the bot asks which unit, which room, whether the unit is making noise or leaking, and whether a photo can be sent. The technician arrives knowing the likely problem and brings the right parts.

Opens a clean ticket. The details go straight into your maintenance system with the right category and priority. Nobody retypes anything.

Keeps the tenant informed. The bot can confirm receipt, share the expected visit window and send an update when the job is closed.

Works in the tenant's language. In a market like the UAE that often means English and Arabic, at any hour.

Where it needs limits

Emergencies. Gas smell, flooding, electrical sparks, no cooling for a vulnerable resident. The bot must spot these words and situations and pass them to a person at once, with a clear message to the tenant about what to do in the meantime.

Promises. The bot should not guarantee times or costs it cannot control. Use language such as "we have scheduled" only when the schedule exists in the system.

Responsibility and billing. Whether a repair is the tenant's cost or the owner's is a policy question. The bot can explain the policy you have written. It should not make exceptions.

A human door. Always offer a way to reach a person. A bot that traps the tenant damages trust more than no bot at all.

What it needs behind it

A bot is only a front door. Behind it you need a working process: someone who assigns, contractors who respond, and a way to close the loop. If those are weak, the bot will make the weakness more visible, faster. Fix the process first, or at the same time.

How to measure it

• Time from first message to a correctly categorised ticket.
• Share of tickets that needed no follow-up questions.
• Time from request to closure.
• Tenant satisfaction after closure.

Learn from the conversations

Read a sample of conversations every week at the start. You will find questions the bot cannot answer, wording tenants use that you did not expect, and process gaps that have nothing to do with the bot. That feedback is often worth as much as the automation itself.