Your tenants do not open the app at 2am. They call.

Ask a landlord where maintenance requests actually come from and almost nobody says the tenant portal. They say a phone call during dinner, a text with a blurry photo, a voicemail on a Saturday. That gap is the whole problem with putting your AI in a chat box.

The assistant is real. It is in the wrong room.

Most property platforms have spent the last two years adding an AI chat panel to a dashboard the tenant has never logged into. The technology works. It is just waiting somewhere the problem does not arrive.

We built Rentari.ai's maintenance assistant around where reports genuinely land, which meant starting with the telephone rather than the browser.

One assistant, four ways in

Luna handles maintenance intake, and it is the same brain behind every channel. A tenant gets the same questions and you get the same quality of ticket no matter how they reached you.

• By phone. Your account gets a dedicated number. Luna answers around the clock, works out what is wrong and which unit, and leaves you a ticket with the transcript attached.
• By text. A tenant texts what happened in their own words. Luna replies, asks the follow-ups she needs, and files the request once the picture is complete. A real back and forth, not a keyword parser.
• In chat. The assistant in the tenant portal offers one safe thing worth trying first, then prepares the request for the tenant to send.
• From a printed link. A QR code for the hallway or the back of a cupboard door. No app, no account, no password, which matters more than it sounds for a tenant who has never signed in.

All four produce the same thing: a drafted ticket, categorised and prioritised, waiting on you. Luna prepares the decision. You make it.

The part we would defend hardest: safety does not run on the model

Before a single word reaches the AI, on every channel, a deterministic check reads the tenant's raw message. It looks for the situations where a wrong answer is not an inconvenience: a gas smell, smoke or fire, a live electrical hazard, a collapse, someone trapped. On a match the tenant gets the emergency instruction word for word, you are paged immediately, and the model is never consulted at all.

It runs in English and Spanish, and it is deliberately narrow. "I smell gas in the kitchen" trips it. "The kitchen tap drips" does not.

The reason it sits above the AI rather than inside it is worth stating plainly. A language model can time out, it can be unavailable, and it can be talked out of its own instructions by a clever prompt. None of that is allowed to stand between a tenant and an evacuation instruction. A safety rule that lives only inside a prompt is a safety rule you are hoping holds.

The same principle governs your settings. You can switch any channel off and you can put the assistant on a schedule. Neither can suppress the emergency path.

New: decide the hours your AI works

The most common request we heard was not "make it smarter". It was "I answer my own phone during office hours, I want help with the evenings".

• All the time. Luna answers everything, day and night.
• Only outside my hours. You set your working hours, days and time zone. During them, calls ring your own phone. Outside them, and all day on a holiday, Luna takes it.
• Only during my hours. The reverse, if you would rather she worked while someone is around to act on what she escalates.

Holidays are handled properly rather than as an afterthought. US federal holidays are recognised, including the weekday a weekend holiday is actually observed, and you can add your own closure dates. A holiday counts as time you are away from your desk, all day.

When a call arrives while Luna is off duty and you have given a number, the caller hears the emergency instruction and is then put through to you. If you do not pick up, or the line is busy, the call falls back to Luna rather than dropping. A missed handoff should never become a missed repair.

How this compares

Almost every platform now markets AI. The useful questions are what it is allowed to do, and where it will actually meet your tenant.

Buildium added AI Insights for reporting in 2024, but it does not prepare a decision. Every screening, lease, renewal and late-rent notice is still yours to draft and send by hand. Pricing starts at $62 a month with no free tier, and incoming EFT carries a per-transaction fee on top.

AppFolio Realm-X is a genuine AI assistant and the most serious of the alternatives here. The catch is who can reach it: a 50-unit minimum and a $400 a month floor, built for property-management firms rather than a landlord who wants AI from day one.

DoorLoop AI summarises reports. It does not draft your screenings, leases, renewals or notices, so the actual work stays manual.

TurboTenant has no AI co-pilot. Screening, leases and reminders are all manual, with nothing to triage maintenance or prepare a decision for you.

Rentari.ai is free forever for unlimited properties and units, with no card to start. Prime is a flat $10 a month for 1 to 5 units, then $2 per additional unit, and you keep 100% of the rent with no per-transaction fee taken off the top. The maintenance assistant answers by phone, text, chat and printed link, keeps a deterministic emergency check above the model, and works the hours you choose.

Competitor pricing and features as of July 2026, taken from each vendor's own published pricing and from third-party review sites. Verify current figures directly with each vendor, as plans and fees change. All product names are trademarks of their respective owners, and Rentari.ai is not affiliated with, endorsed by, or sponsored by any of them.

What it does not do

Worth being straight about, because "AI property manager" is doing a lot of work in most marketing.

Luna does not send a contractor to your property on her own by default. She picks the vendor she thinks fits, writes the job up, and waits for your approval. If you want her to book within limits you set, that is a setting you switch on deliberately, with a spend cap and an acceptance window, and an emergency is never dispatched automatically no matter what those limits say.

She also does not guess. Asked something she has no grounding for, she says she does not know and routes it to you rather than inventing a policy, a fee, or a date.

Getting started

Maintenance intake is on by default, so a new account already answers chat and the printed link. Adding a phone line takes about a minute from the dashboard, and setting your hours takes about the same.