AI property management software describes a range from a spam filter on the maintenance inbox to a system proposing which applicant to house. The label tells you nothing about which, and the difference matters because the risk is completely different at each end. This page is four questions that put a product somewhere on that range before you evaluate anything else.
What does it decide, and what does it merely draft
Drafting is low risk: a suggested reply to a maintenance message, a first pass at a listing description, a summary of a long thread. Deciding is high risk: scoring an applicant, setting a rent, flagging a tenant. A product that drafts and leaves the decision with you needs far less scrutiny than one that decides, and vendors often describe both in the same sentence. Ask which it is, feature by feature.
What happens when it is wrong
Not whether it is ever wrong, which it is, but what the product does about it. Can you see what it based an output on, override it, and is the override recorded. A tool whose reasoning is invisible and whose outputs are not attributable is one you cannot explain to anyone later, and explaining a decision later is a large part of what a landlord's records are for.
Anything touching applicants is a compliance question first
Screening and selection sit under federal fair-housing law and, where a consumer report is involved, under the rules on permissible purpose and adverse action. An automated score that contributes to a refusal does not change what you owe the applicant, and a model that produces a disparate outcome is your problem rather than the vendor's. That is a reason to keep selection decisions human and documented.
Where the data goes
Tenant names, addresses, payment histories and messages are personal data, and a feature that sends them to a third-party model is a processing arrangement whether or not anyone called it that. Ask what leaves your tenant, what is retained, whether it trains anything, and get the answer in the contract rather than in a sales call. Then tell your tenants what you have agreed, because they are the subject of it.
Questions people ask about ai property management software
Is AI in property management worth having?
The drafting and summarising end saves real time at low risk. The deciding end buys less than it appears to and brings compliance exposure a small operator is poorly placed to carry.
Does Lettza use AI to decide anything?
No. This hub's worksheets are arithmetic over figures you enter, and the record keeps what you put in it. There is no scoring of applicants and no screening of any kind.
What should I insist on contractually?
That personal data is not used to train anything, that it is deleted on termination, and that you can export the full record. Those three make a vendor decision reversible.