"AI agent" is a vague phrase until you watch one do a day's work. Here is a Tuesday for a manager with twenty-eight units, with Tensor Autonomous running three responsibilities from the Library. Everything below is what the starters do as published; the names are made up.
7:40 — three inquiries from overnight
Two from the listing site, one forwarded by the owner. You paste the three into Reply to rental inquiries (with Start from new email turned on, they would have started it as they arrived); it reads each against your listing facts and proposes a reply, which you approve into your drafts:
Hi Jordan, thanks for asking about the 2-bedroom on Maple Row. It is available from November 1, and a 30-pound dog is welcome under the pet terms in the lease. Parking is one space. Showings are on weekday evenings between 4 and 7; would Wednesday at 5 or Thursday at 6 suit you?
You read the three drafts over coffee, change "Thursday at 6" to "Thursday at 6:30" on one, and send them. Tensor sent nothing; it only wrote.
9:15 — the leaking sink
A tenant emails: water under the kitchen cabinet, a bucket in place, home all day Tuesday. You paste it into Log maintenance requests; it proposes the row for your maintenance log with the unit, the issue and its urgency (an active leak is an emergency), and the reply:
Hi Priya, first, if you can reach it safely, turn off the valve under the sink and keep the bucket in place. I have logged the leak under the kitchen sink in 2C as an emergency. I will confirm with you before anyone comes to the unit, and I will be in touch about timing once I have spoken to the plumber.
You send it and call the plumber. The log now shows the request, when you logged it and that the tenant was told. The plumber's visit is yours to arrange; recording it is a row you fill in, or a responsibility you describe in your own words.
11:00 — a showing request
"Can I see unit 2A next Thursday at 2pm?" You hand it to Book showings; it reads Thursday on your calendars, finds 2:00 free and inside your showing hours, and proposes one 30-minute showing. The proposal sits in front of you with the time, the unit and the prospect's name. You approve it; the event appears on the calendar, Tensor reads it back, and the receipt shows it matched. Nobody is invited and nothing else on the calendar moves.
Had 2:00 been taken, the starter would have said so rather than guessing another time. Had the request been for Sunday, outside your hours, it would have booked nothing and told you why.
2:30 — the one it stops on
A fourth inquiry asks whether you would take $200 off the rent for a two-year lease. There is no listing fact for that, so the draft Tensor leaves says you will come back to them on the terms, and the question is in your hands. That is the agent working as it should: it writes what it knows and stops where it doesn't.
5:10 — the receipts
Five runs today. Each has a receipt: the email or request it read, the row or event it wrote, the draft it saved, read back and compared, and who approved what. If the owner asks on Friday why unit 2A is showing Thursday at 2, the answer is one click away.
What made this a good day
- Nothing went to a tenant that you had not read. The inquiry and maintenance starters draft; you send.
- The one write to a calendar waited for your OK. Later, if you like, you set limits (Tensor may book without asking on weekday mornings, up to three a day) and that wait goes away for the routine ones. Sending email, and moving or cancelling an event, always ask.
- Nothing was invented. Prices, availability and pets came from your listing facts; the one question outside them came to you.
What it is not
An AI agent in this sense is not property-management software, and it does not replace yours: leases, rent, accounting and official work orders stay where they are. It is not an answering service; it works in your Gmail and Google Calendar, today, with Microsoft 365 to come. And it is not autonomous by default. It becomes more autonomous only as you set limits for the routines you have watched it get right.
Where to start
- Tensor for property management: the three starters and their sample requests.
- Try it live: a rental inquiry on sample data, approved by you.
- Plans from $149 a month.
Related reading
- Property management automation for small portfolios
- A maintenance request log that keeps tenants and vendors moving
- What is an AI agent?
See it in a demo
Watch one of these runs on sample data, approval and receipt included.