An AI agent for real estate: new listings, new owners and a daily note
What an AI employee can watch for an agency each morning: new listings, new owners in county records, and a short note to each agent. Where the data exists.
Part of the guide AI employees
An AI agent for real estate is most useful for the watching an agency never has time for: which homes were listed overnight in the streets you work, which properties changed hands in the county records, and a short note to each agent saying what to look at first. In xMap Flows that is a scheduled AI employee using Regional data (property listings), Property changes (US county records), xMap LLM to write the note, and Email or Slack to deliver it before the morning meeting.
Be clear about coverage before you build. Property listings are offered in the United States, Japan and the United Arab Emirates. Property changes, meaning new owners, sales with their price and changed values, come from public property records of some US counties only. Outside those places, the same idea still works with xMap's places database (new businesses and buildings around your patch), but not with listings or ownership.
Three jobs, three flows
| Job | Main block | What each item holds | Where it works |
|---|---|---|---|
| New listings in my patch | Regional data, Property listings | Type, rent or sale, price, size, address, when listed or updated | United States, Japan, UAE |
| New owners and recent sales | Property changes | Address, new and previous owner, sale price, date | Some US counties |
| A note for each agent | xMap LLM, the whole list | A few lines on what matters today | Anywhere, on the items above |
Keep the three separate at first; combine them once each is useful on its own.
New listings each morning
Pick the country and the Property listings dataset in the Regional data block, then where to look (a point and a distance, or the areas of an Area of interest before it), rent or sale, the kind of property, and price or size limits. Put Only new after it, so each listing reaches you once. A Filter on the price per square metre, or an xMap LLM score against what your buyers asked for, narrows it further.
In Tokyo the listings also carry the nearest station and the walk to it, which buyers there ask about first. Use it in the note.
New owners from county records
Property changes reads a county's public property records each run and brings in what changed since the last look: new owners, sales with their price, new values, new uses. Pick the county from the list (only counties with updates are offered) and the kinds of change you care about. A new owner is a lead for a lot of trades, and for an agency it is also a signal about which streets are moving.
The daily note
The note is where the AI earns its keep. Set xMap LLM to work on The whole list and give it a plain brief:
- Say who it is for: a residential agent covering Coral Gables and South Miami.
- Say what to pick out: listings under the asking range of our open buyers, homes sold within two streets of our current listings, anything that went back on the market.
- Say what to leave out: do not guess values; quote only prices in the data.
- End with an Email to the agent, or the Slack app block to the team's channel, so it is read on the phone before the first viewing.
A worked example (made up)
A five-agent brokerage in Miami covers three neighbourhoods. The office manager builds two flows. The first, every weekday at 6:30, uses Property listings for homes for sale within 2 km of each neighbourhood's centre, Only new, and an xMap LLM score against a short note on what their buyers want. The second reads Property changes for Miami-Dade, keeps sales and new owners in their three neighbourhoods, and passes them on.
A third flow, at 7:15, uses Team findings to read what the other two found and has xMap LLM write one note per agent, split by neighbourhood. The first week, one agent notices that two homes on her street sold for well under asking, and calls the owners of a third that is still listed. Nothing in the flow told her to; the note simply put the three side by side.
Mistakes to avoid
- Expecting listings or ownership data outside the places listed above. Check coverage first.
- Asking the AI for valuations. It writes from the listings and records it was given, not from a model of prices.
- One giant flow. When something breaks, three small ones are easier to fix.
- Sending every listing to every agent. Split by area, or the note becomes noise by Wednesday.
For buyers, a chat page can answer questions about a neighbourhood from live data; the neighbourhood guide chat template starts one. The AI employees guide explains how such employees report back and how much they may do on their own.
Questions
Which countries have property listings?
The United States, Japan and the United Arab Emirates. Listings say only that they come from a property site; the sites are not named.
Does it show who owns any home?
In US counties that xMap's property service covers, Property changes can bring the owner of the parcel at each of your places, or every change in a county. Elsewhere there is no ownership data.
Can it send listings to buyers directly?
It can email them, but most agencies keep the agent in the loop: put an Approval before any message that leaves the agency.
Will the same listing come up every day?
Not with Only new after the listings block. A listing that is updated keeps its identity, so it is not sent again as new.
Templates that do this
Read next
More on ai employees and automation · Every article · Templates