AI for real estate agencies: portal leads answered in minutes, viewings booked and paperwork chased for you

A new flat goes live on the portals on Tuesday at 8 pm and by 10 pm you have fourteen enquiries. Half ask whether it’s still available, three want to view on Saturday, two haven’t sorted their mortgage and one writes in German from Munich. Meanwhile you’re showing another property, the seller is messaging you to ask how it’s going, and there’s a rental file on your desk still missing two payslips. This page explains which parts of that work AI can do for a real estate agency, which parts should stay with your agents, and what it costs.

We’re based in Spain, so the examples use Spanish portals (Idealista, Fotocasa, Habitaclia) and Spanish paperwork, but the same setup works with Rightmove, Zoopla, Immowelt or any portal that sends you leads by email or into your CRM.

The short answer

• What you can automate: replying within minutes, at any hour, to enquiries from property portals, your website or Instagram; qualifying them on WhatsApp or email; matching buyers with your listings and alerting them; booking viewings in each agent’s calendar with confirmations and reminders; collecting and checking documents for sales and lettings; writing and translating listing descriptions; drafting valuations and weekly seller reports.
• What it shouldn’t do: choose the tenant, set the asking price, make up property details or answer legal and tax questions. Those stay with you.
• What it saves: hours of copy-and-paste replies, leads that go cold because nobody answered at the weekend, viewings with people who were never going to qualify, and days spent chasing paperwork.
• What it costs: an assistant that answers from your listings, from around €1,000; an agent connected to your CRM and calendars (1–3 tools), €2,000–5,000; multi-system or phone (voice) agents, from €5,000. Plus usage, which is cents per conversation. Details in our guide to how much an AI agent costs.
• How to start: with portal leads: instant reply, qualification and a booked viewing. It’s where the difference shows first and the easiest thing to measure.

What you can automate in a real estate agency

Process	What AI or automation does	Who benefits
Portal and website leads	Reads the lead from the portal or website, replies straight away on WhatsApp or email and asks your qualifying questions	Sales agents, buyers, tenants
Buyer–listing matching	Compares each buyer’s search with your stock and alerts them when something suitable comes in	Agents, buyers
Viewings	Offers real slots from the agent’s calendar, books, confirms, reminds and handles cancellations	Agents, clients
Listings and translations	Writes the description from the property sheet and photos, and translates it for overseas buyers	Listings team, marketing
Photos and virtual staging	Edits photos and creates virtually furnished versions, always labelled as such	Marketing
Documents	Requests and checks ID, payslips, employment contracts, land registry extracts or energy certificates, and chases what’s missing	Admin, agents
Valuations	Drafts a report from comparables for a person to review and sign	Valuers, managers
Seller reports	Summarises enquiries, viewings and feedback for each property every week	Sellers, agents
Follow-ups and reviews	Messages after viewings and completion, and invites every client to leave a review	Management, marketing
CRM hygiene	Flags duplicate contacts, incomplete records and leads with no owner	Management, the whole team

Portal leads answered in minutes, even on Sunday

Before: the portal notification lands in your inbox or in your CRM (Inmovilla, Witei, Mobilia or whatever you use). If the agent is out on a viewing, they’ll see it in an hour; if it’s Sunday, on Monday. By the time they reply, the enquirer has already spoken to two other agencies. And many conversations start the same way: “Is it still available?”, “Are pets allowed?”, “What are the service charges?”

After: the AI agent reads the lead, identifies the property from its reference and messages the enquirer on WhatsApp or email within minutes. It introduces itself as the agency’s AI assistant, answers from the real property record (price, size, community fees, parking, availability) and asks the questions you choose: for buyers, whether they have a mortgage in principle or a property to sell; for tenants, move-in date, number of occupants and whether they can provide proof of income. Everything goes into the contact record in your CRM with a summary, and the agent only gets the people who are ready to view.

Just as important is what it doesn’t do: if someone asks something that isn’t in the property record (“Could we add another floor?”, “How much tax will I pay?”), it doesn’t guess. It says a colleague will reply and flags it for a person.

Matching buyers with listings and alerting them

Before: buyer requirements live in the CRM and in each agent’s head. When a new property comes in, someone remembers two or three buyers and messages them; everyone else finds out from the portal, or not at all.

After: every time you add a property, the system matches it against active searches (area, budget, bedrooms, lift, terrace) and sends a personalised message to the buyers who fit, with the link and the option to book a viewing in the same chat. Older contacts are first asked whether they’re still looking, so your CRM cleans itself up and you stop messaging people who have already bought.

Viewings booked straight into each agent’s calendar

Before: five messages to agree a time, another to confirm the address and a missed call on Saturday from someone who can’t find the building. And there’s always someone who doesn’t turn up.

After: the AI agent checks real availability in the responsible agent’s calendar (Google Calendar, Outlook or the CRM’s own), offers two or three options, books the one the client picks and sends a confirmation with the address and who will meet them. The day before, it sends a reminder with buttons to confirm or reschedule. If someone cancels, the slot goes to the next interested person for the same property. For tenanted homes, it only offers the times the tenant or owner has agreed to.

Listing descriptions and translations for overseas buyers

Before: descriptions are written in a hurry or copied from the last one, and the English or German version, if there is one, comes from a machine translator nobody checks.

After: AI writes the description from the property sheet, the photos and your notes from the valuation visit, in your agency’s tone, and adapts it into English, German, French or whatever your buyers speak. An agent reviews it before it goes live. One firm rule: the text can only say what’s in the property sheet. If the AI writes “newly renovated” or “sea views” and that’s not on record, that’s your liability, not the tool’s. Mandatory information (in Spain, for example, the energy rating) should come from the property data, not from the AI’s prose.

Photo editing and virtual staging, honestly

AI can fix lighting, straighten photos and show what the empty living room would look like furnished. That’s useful for empty homes, but there’s a limit: an image must never hide defects or change what the buyer will find. Always label virtually staged photos (“virtually staged” or “digitally furnished”) and publish the real photo as well. A buyer who doesn’t recognise the flat at the viewing is a wasted viewing and, at worst, a complaint about misleading advertising.

Documents for sales and lettings

Before: you ask on WhatsApp for ID, the last three payslips, the employment contract and the employment history. They arrive as crooked photos, across five messages and three days, and someone has to work out what’s still missing.

After: the client gets a link to a secure form where each document has its own slot. AI checks that each file is what it claims to be, that it’s legible and in date, pulls out the data you need for the file and sends a friendly reminder for anything missing. On a sale it does the same with the seller: in Spain, the nota simple (land registry extract), the energy performance certificate, the latest property tax (IBI) receipt and the community certificate confirming there are no outstanding fees. Your team reviews a complete file instead of chasing paper. The technical side is covered in our document and image automation service.

Valuation drafts from comparables

Before: to prepare for a valuation visit, someone searches the portals by hand for comparables, pastes them into a spreadsheet and writes the report.

After: AI starts from your own completed sales and the sources you’re allowed to use, picks comparables by area, size and condition, and drafts a report with a price range and the reasoning. The agent adjusts it with what only they know (the street, the light, the neighbour on the fifth floor) and signs it. This is a market appraisal, not a formal valuation: in Spain, valuations for mortgages are done by approved valuation companies.

Weekly seller reports

Before: the seller rings every week to ask how it’s going, and the answer depends on what the agent remembers.

After: every Monday a short report goes out with enquiries received, viewings done, what viewers said and, if needed, a suggested price adjustment that the agent approves before sending. Fewer calls, and the price conversation is based on data.

Follow-ups and reviews

After each viewing, a message asks the viewer what they thought; the answer feeds the seller report. After completion, every client is invited to leave a Google review, not only the ones you think are happy: Google’s policies prohibit selectively asking for positive reviews, and a complaint that arrives through this channel is a chance to call the client before they write it publicly.

CRM hygiene

An agency CRM fills up with duplicate contacts, numbers with no name, leads assigned to agents who left months ago and searches nobody updates. A weekly automation flags duplicates, fills in records with what was said on WhatsApp and highlights leads with no owner. It isn’t glamorous, but everything above works better on a clean CRM.

What a typical rollout looks like

1. Map the lead journey (one or two workshops). Where enquiries come from, how they’re shared out between agents, which questions actually qualify a buyer or tenant in your area, and which CRM and calendars you use. Much of the value is in writing this down.
2. First process: portal leads and viewings, with a person approving the agent’s messages for the first few weeks. We save every conversation and review the ones that go wrong.
3. Give autonomy where it’s earned. We use Jev, a decision model from TypeSafe that returns calibrated probabilities instead of text, to decide whether a reply can go out on its own or needs a person: “Yes, it’s still available, does Thursday at 6 pm suit you?” can be sent automatically; a price negotiation, a complaint or a question about the contract goes to an agent.
4. Second process: documents and files, reusing the same channel and the same CRM.
5. Then the rest: buyer matching, listings, seller reports and valuations, each as a measurable step.

What still needs a person: winning instructions, the viewing itself, negotiation, choosing the buyer or tenant, pricing and anything legal. AI takes the repetitive replies, the diary, the chasing and the first drafts.

The risks of doing it alone

A real estate chatbot built in an afternoon works in the demo. The problems show up with real clients:

• Buyers’ and tenants’ personal data. Payslips, ID, employment history and family details pasted into a free chatbot or sent to a model that stores them or trains on them. Under GDPR your agency is the controller: you need a lawful basis for each piece of data, should only ask for what you need, must delete the files of applicants who didn’t rent, and need a processing agreement with every provider. For EU clients we use models processed in Europe, with zero data retention and no training; see our guide to EU data residency for AI models and GDPR.
• Portal data. The leads a portal sends you are yours to answer that enquiry, not for any purpose you like. And automatically extracting listings from portals for valuations or prospecting can breach their terms of use and the data protection rights of private sellers. Check them before you automate.
• Wrong or invented property details. A badly connected agent that says the flat has a lift, the fees are €40 a month or pets are allowed when they aren’t produces wasted viewings and angry clients. The agent must answer only from the property record, and the record has to be right.
• Discrimination in tenant screening. An automatic filter that rejects people by nationality, surname, age or children is illegal: in Spain, Law 15/2022 on equal treatment prohibits discrimination in access to housing, and fair housing rules exist in most countries. GDPR Article 22 also restricts decisions based solely on automated processing that significantly affect someone. The agent can collect and organise documents; a person decides, using objective criteria.
• Messages to the wrong client. Sending one buyer’s offer to another, or one seller’s report to someone else, is a data breach as well as a commercial problem. Every conversation has to be strictly tied to its contact and property, and tested on real cases before launch.
• Your agency’s WhatsApp number. Connecting a regular number (Evolution API) has no per-message cost and works well for one-to-one conversations; the official WhatsApp Business API is the route for high volumes and bulk messages, such as sending new listings to all your buyers. If the number you use with clients gets banned, you lose every conversation in progress. We compare the two in Evolution API vs WhatsApp Business API.
• Silent failures. The CRM integration stops working on a Friday and weekend leads get no reply. Every agent we build logs every conversation, and failures trigger alerts.

An experienced partner doesn’t make these risks disappear, but knows where they are before the first client writes in.

Want to see how it would work in your agency? Book a free call and tell us how your leads arrive today.

How we do it at Aitomatik

• We start from your process, not a tool. If your property CRM already does something well, we connect to it rather than replace it. When nothing on the market fits, we build it: see our custom software development with AI service.
• Agents in production, not mock-ups. For a European platform we built a multilingual phone agent and chatbot connected to their systems that resolve 80% of enquiries without human involvement, around the clock. It’s the same pattern an agency needs for leads that come in on a Sunday.
• WhatsApp chosen per project. We use the official API or Evolution API depending on volume and risk. In a project we’re building for a domestic staffing agency, the WhatsApp agent qualifies candidates with no per-message cost, a problem very close to qualifying tenants.
• Prospecting and data. If you also want to find sellers or enrich your contact base, see our lead generation and data enrichment service.
• Every conversation saved and evaluated. We test every change against real conversations before shipping it. A dashboard where your team can read conversations and take over a chat is optional.
• Clear pricing. The tiers above come from our guide to how much an AI agent costs. If you’re new to the topic, start with what an AI agent is.

We’re open to projects with estate agencies, agency networks and developers, in Spain and beyond. The examples on this page are what we would build, not a list of clients in the sector. If you manage rentals or buildings, see also AI for property management, and browse all our guides.

Frequently asked questions

What is the best AI for real estate agents? It depends on the job. For writing listings and emails, ChatGPT, Claude or Gemini on a business plan work well. For answering leads, booking viewings and collecting documents you need something connected to your CRM, your calendars and WhatsApp: an AI agent, either the one built into your CRM or a custom one. We compare the general assistants in ChatGPT vs Claude vs Gemini for business.

Is there a ChatGPT for real estate agents? Not an official one, but you can use ChatGPT or any other assistant for drafting. The limit is that a general chatbot can’t see your listings, your CRM or your diary, and you shouldn’t paste clients’ payslips or IDs into a free version. That’s the gap a connected AI agent fills.

How much does an AI chatbot for a real estate agency cost? From around €1,000 for an assistant that answers from your listings; €2,000–5,000 for one connected to your CRM and calendars; from €5,000 if it joins several systems or answers the phone. On top of that, cents per conversation in AI usage and, with the official WhatsApp API, Meta’s per-message fees.

Will AI replace real estate agents? No. It replaces repeated answers, diary juggling and chasing paperwork. Winning instructions, viewings, negotiation and the seller’s trust stay with the agent, who gets more time for them.

Can I use AI to screen tenants? You can use it to collect documents, check they’re complete and organise applications. The decision should be made by a person using objective affordability criteria, with no filtering by nationality, origin, age or family, and asking only for the data you need.

How can real estate agents use AI day to day? Start with the tasks that eat time without needing judgement: replying to portal enquiries, booking and confirming viewings, chasing documents and writing first drafts of listings and seller reports. Keep the decisions with people.

Let’s start with your portal leads

Tell us where your enquiries come from, which CRM you use and how long it takes you to reply today. We’ll tell you what we’d automate first and what it would cost. See our AI customer service service or book a free call.

Indicative prices for SME projects as of October 2026. Legal references are for information only and aren’t legal advice. Last reviewed: 15 October 2026.