Learn by doing
AI for estate agents: where it helps and where it doesn’t.
AI is good at drafting and sorting. It shouldn’t set prices, invent property details or send messages nobody has read.
By Max O’Sullivan · Guide and worksheet
Where AI helps an agency today
- First replies to portal enquiries. A draft answer built from the listing’s facts, ready for an agent to check and send.
- Listing descriptions. Turning a fact sheet into a readable description, in English and Spanish.
- Translation. A quick first version for a buyer who writes in another language, checked by someone who speaks it.
- Notes after a viewing or call. A short summary with the next actions and who owns them.
- Sorting incoming messages. Buyer, seller, rental or viewing request, so each one reaches the right person.
Not every one of these needs AI. If a task follows fixed rules, plain automation is cheaper and more reliable. The guide AI or automation explains the difference.
Where it goes wrong
- It invents details. Sea views, square metres or a walk to the beach that aren’t in the listing.
- It sounds sure when it’s wrong. Especially on tax, legal or mortgage questions. Those answers should come from a professional.
- Personal data goes somewhere. Buyer names, phone numbers and documents are covered by data protection law (the GDPR, or RGPD in Spain). Check where a tool sends data before pasting anything in, and never paste ID documents.
- The tone drifts. A reply sent without reading can be too pushy, too casual or simply wrong for that buyer.
Keep a person in the loop
Give the AI an approved list of facts for each property, and ask it to use only those. Treat every result as a draft. An agent reads it, corrects it and sends it. Keep a note of what had to change, so you can see whether the drafts are actually saving time.
Before you rely on any of this day to day, test it with real-looking cases, including awkward ones. The guide how to test an automation shows how.
Worked example: a reply that added a sea view
Fictional training example — not a client project.
A buyer writes in English through a portal, asking about a townhouse in Estepona. The agent pastes the listing’s fact sheet into an AI tool and asks for a short, friendly reply that offers two viewing times.
The draft is good, but it says the house is “a five-minute walk to the beach”. That isn’t in the fact sheet. The agent removes it, checks the viewing times against the calendar and sends the reply. The next day they add a line to the instructions: “Use only the facts provided. If something is missing, leave it out.”
Exercise: choose one task to try with AI
Pick one piece of writing your agency repeats every week. Copy these prompts into your notes, and keep real customer details out of the exercise.
Your working notes
Work through one task at a time.
- The task
What do you write again and again, and how often?
- The facts
Where do the facts come from, and who keeps them up to date?
- The limits
What must the draft never say or promise?
- The check
Who reads each draft before it’s sent?
- The data
Does the task involve personal data? Which tool would see it?
- The test
How will you know after two weeks whether it saves time?
When it’s worth building something
Copying and pasting into a chat window is fine for trying an idea. If the same task happens every day, it’s worth connecting your tools so the draft appears where your team already works, with the facts filled in.
That’s the kind of system I build for estate agencies, with a person checking every message. See websites and systems for estate agencies.