Where AI Should (and Shouldn't) Run Your Property Operations
The useful question is not “will AI change property management.” It is which tickets a model may close without a person, and which ones it may only prepare.
Use this playbook to pick one workflow, write the guardrails, and measure whether speed actually improved.
Decide what AI is allowed to do
Write a one-page policy before you turn anything on. If staff cannot point to the rule, do not automate the action.
| Action | AI may do this | Human must do this |
|---|---|---|
| Intake | Ask for photos, access notes, and “when did it start?” | Override when the resident is distressed or the unit is occupied by a vulnerable household |
| Classify | Tag plumbing / HVAC / electrical / pest and suggest urgency | Declare emergency (gas, active flood, sparking electrical, no heat in extreme cold) |
| Route | Apply after-hours and vendor rules you already published | Approve a new vendor, overtime, or exception to the playbook |
| Message | Send “we received this” and the next update time | Handle disputes, accommodations, and anything that sounds like a legal notice |
| Close | Draft a work-order summary | QA the vendor and decide if the ticket is actually done |
If a step is not in the “AI may” column, it stays a draft until a person clicks approve.
Week 1: pick one high-volume workflow
Choose the inbox that already has a written SOP. Good first candidates:
- Maintenance intake — same questions every time, photos required, emergency list published.
- Resident status updates — “received / scheduled / delayed / complete” with a next-update timestamp.
- Lead-to-tour routing — availability and pre-qualification questions, not screening decisions.
Skip pricing, lease drafting, and collections automation until intake is boring and reliable.
Measure a two-week baseline before go-live:
- Median time from request to first human-quality acknowledgment
- Share of tickets that bounce back within 30 days
- Hours coordinators spend asking for the same three facts
The intake script that actually helps
A useful assistant does not “troubleshoot 24/7.” It refuses to route until the record is complete:
- Unit and access instructions
- Photo or short video of the issue
- Whether water is off, power is on, and anyone is without heat or cooling
- Preferred access window
Then it applies your urgency list. Publish that list in the tenant portal so residents see the same rules staff use. Align emergency definitions with habitability and safety guidance (many teams start from public HUD fair housing and housing quality resources).
Where predictions earn their keep
Skip “the HVAC will fail next Tuesday” until you can act on a simpler forecast:
- Which buildings generate the most repeat tickets
- Which vendors reopen the same unit within 30 days
- Which equipment categories spike after a weather event
Those are operational queues, not magic. Put them on a weekly review, not a dashboard nobody opens.
Money and compliance stay human
AI can flag duplicate invoices, odd utility spikes, or missing receipt attachments. It should not post adjustments, issue refunds, or change rent without an approver. Same rule for screening and accommodations: consistent criteria, documented by staff.
How to adopt this without a second inbox
- Connect the assistant to the work-order and message thread that staff already use.
- Require an approval step for dispatch, after-hours send, and any financial action.
- Review ten random tickets every Friday for invented facts or skipped safety questions.
- Expand only when acknowledgment time and rework both move the right direction.
Unitz can run that intake-and-routing loop inside the same maintenance and messaging workspace your team already uses. If you want to walk one portfolio workflow before you turn it on, book a consultation.
Frequently Asked Questions
Related articles
Share this article