Calling the listings
Dozens of agencies call the same private seller. Each call meets a more saturated owner.
PropTech · Artificial intelligence
ATLAS finds the properties that are about to be sold before they are listed.
An artificial intelligence platform for real estate agencies. It reads public records, official gazettes and listing portals, works out which properties are likely to be sold and assigns each opportunity to the right agency, always with the owner’s consent.
Status In development · completion Dec 2027 · looking for pilot agencies and investors
The origin
It started on a napkin, in November 2024.
As a consultant at Engel & Völkers in Madrid, one question kept me up at night: how do I talk to a homeowner before they decide to sell on their own through a portal? By the time the listing appeared, the conversation came too late. After many months looking for an answer, I sketched on a napkin a system that would spot owners intending to sell, so we could offer them advice before they published anything, always with their consent. That is how ATLAS was born. Caelum was founded in July 2025 to turn that idea into a company. Fernando Stimac Jaén, founder.
The problem
An agency’s rival is not another agency: it is the owner who thinks selling alone is easy.
Almost everyone starts the same way: they list on Idealista or Fotocasa, often with a “no agencies” note. That is where the trouble begins, and by the time they ask for help the conversation comes too late.
- Lists alonePosts the listing on a portal, convinced that selling or renting is simple.
- The browsers arriveMessages from people who don’t know what they want, unscreened viewings and applicants who don’t meet the requirements.
- Burns outWeeks of calls and viewings with no result. The price drops, and so does the patience.
- Gives inHands the home to the first agency that calls about the listing, or ends up with a bad deal or a bad tenant.
How agencies source today
The same sources for everyone, and they yield less every year.
Neighbourhood farming
Walking streets, mailboxes and buildings: slow, costly and hard to scale.
Information at a price
Some neighbourhood information only arrives in exchange for an informal commission, with no guarantees.
The owner app
Real advice before anything is published.
Owners sign up and we check that they own the home. That way we know the person asking for the valuation is the owner, and they decide whether to talk to an agency.
Verified owner
We check that the person requesting the assessment is the owner.
Land registry extract
The registry information for their home, requested from the app itself.
AI condition check
AI reviews the photos and estimates the condition of the home.
A data-backed price
A valuation based on real market data, not on gut feeling.
How it works
From a public signal to an owner asking for help.
- CollectionContinuous reading of official public sources and market data.
- AI interpretationModels that read legal texts, cross-check sources and remove duplicates.
- ScoringAn intent-to-sell score from 0 to 100, calculated with fixed, traceable rules.
- AssignmentThe opportunity goes to the right agency only once there is consent to make contact.
Architecture
Three AI agents, each with one job.
Separating the roles makes every decision reviewable: who detected the signal, why it scored as it did and why it reached that agency.
Detects and scores
Gathers signals from public sources and calculates each property’s intent-to-sell score.
Decides where it goes
Assigns each opportunity to a subscribed agency. It allows no contact without the owner’s verified authorisation.
Supports the agency
It is the estate agent’s interface and learns from what happens to each opportunity to sharpen the next ones.
Who it is for
Two sides of the same market, each with a clear reason to take part.
Agencies
Real estate agencies
A monthly subscription that delivers prioritised opportunities in their area before they reach the open market.
- A shortlist of properties likely to be sold.
- The reason behind each score, signal by signal.
- Contact only with owners who have authorised it.
Owners
Owners
A free valuation, the land registry extract and an AI condition check of their home. Then the owner decides whether they want an agency’s help.
- No-obligation valuation.
- Explicit, revocable consent.
- No sales contact without their permission.
Compliance
Designed to withstand an audit.
ATLAS is designed in line with the General Data Protection Regulation and the EU Artificial Intelligence Act.
Contact only with consent
Public signals are used to prioritise. No owner is contacted without verified authorisation.
Explainable scoring
AI interprets the documents. The final score comes from fixed rules, so every result can be explained and reviewed.
Every step is logged
Each opportunity records which agent detected the signal, why it scored as it did and which agency it reached.
Roadmap
From research to December 2027
Scroll to follow the project.
Mar 2025
ATLAS · ResearchStudy of real estate sourcing in Madrid: interviews with agencies and analysis of how a property reaches the portals today.
May 2026
ATLAS · DevelopmentScout engine in validation with real public data.
Apr 2027
ATLAS · PlannedCommercial beta: B2B plans for agencies and an app for owners.
2025 — 2027
Oct 2025
ATLAS · ArchitectureDesign of the Scout, Director and Consultor agents, and of the consent framework that decides when contact is allowed.
Oct 2026
ATLAS · TodayWorking prototype and selection of pilot agencies in Madrid.
Dec 2027
ATLAS · GoalDevelopment complete. ATLAS goes into commercial operation with its AI Act documentation in place.
An illustrative overview of research and development work. Milestones after October 2026 are forecasts and may change.
Do you run an agency in Madrid or want to invest in ATLAS?
Let’s talk about a pilot with your agency or the €250,000 seed round.