RealAnalytica
Back to all articles
Real Estate TechnologyOntologyUnified Data LayerKnowledge GraphAI AgentsMCPReal Estate DataReal Estate CRM

What Is a Real Estate Ontology? The Data Layer Behind AI Agents

November 18, 2026
7 min read
What Is a Real Estate Ontology? The Data Layer Behind AI Agents

A real estate ontology is the list of things a real estate business keeps records of, and how those things link: people, homes, listings, deals, tasks and conversations. It is the layer that lets an AI agent answer a question about your business from your records. Some teams call the same idea a unified data layer or a knowledge graph.

This guide explains the term in plain words, shows why AI agents need one, and lists the 16 record types in the ontology we run at RealAnalytica.

What Is an Ontology, in Plain Words?

An ontology has three parts:

  • Nouns. The kinds of record the business keeps: a contact, a listing, a transaction, a task.
  • Attributes. What each record holds. A contact has a name, a stage, a source and a last-contacted date. A transaction has a stage, dates and people.
  • Links. How one record points to another. A task belongs to a contact or a transaction. A conversation is with a contact. A transaction points to a listing.

A spreadsheet of contacts has the first two parts. The links are what it lacks, and the links are what most real questions need. "Is Priya's deal on track, and what did she say on our last call?" touches a contact, a call, a transaction and its tasks. Only the links tie those four records to one person.

Ontology, Unified Data Layer, Knowledge Graph: What Is the Difference?

Term What it means How it relates to an ontology
Ontology The list of record types, their attributes and their links The definition itself
Unified data layer One place where data from several tools sits in one shape The ontology is what gives that layer its shape
Knowledge graph Records and links stored so you can walk from one record to the next The ontology is its plan: which records and which links can exist
RESO Data Dictionary The industry's shared field names for MLS data A standard for exchanging MLS data. A brokerage ontology also has to cover the daily work around a listing: stages, deals, tasks and conversations
MCP server The connector an AI app like Claude or ChatGPT uses to reach a data source The way the ontology is served to an AI app

Why Do AI Agents Need an Ontology?

A list of actions is not enough

We learned this from our own product. Atlas, the AI agents inside RealAnalytica, can take about 175 kinds of action: send an email, update a contact, create a task, change a deal's stage. Each action describes a process. So an agent could know what it was able to do and still have no picture of what exists. It could not say what a deal contains.

The fix was a list of nouns. Once an agent can ask "what kinds of record are there, and what does each one hold?", it can plan a question the way a person would: find the contact, follow the link to her deal, read the open tasks.

A short tool list beats a long one

One design we weighed gave every record type its own set of tools, about 60 in all. We turned it down. Every new record type would have added more tools, and an AI app would have had to read a long list before doing anything. With an ontology, six read tools cover every type: list the record types, describe one, find records, open one record, follow a link, and search by words. A new record type needs no new tool.

Permission follows the nouns

When the business is a list of record types, a person can grant access one type at a time: yes to contacts and tasks, no to email. Changes work the other way round. Each change is its own named tool, because AI apps ask for approval by tool name. One tool for every kind of change would have meant one approval covering all of them.

Every record keeps its source

A contact synced from Follow Up Boss and a contact typed in by hand are both contacts. Each keeps the name of the tool it came from, so nothing is lost when the data lands in one shape.

The 16 Record Types in RealAnalytica's Ontology

We keep the whole ontology in one file beside the database. Every attribute names the column that stores it, so a renamed column breaks the build before it can break an answer. The connector an AI app uses is built from the same file.

Record type What it is Links to
ContactA person in your CRMTransaction, Listing, User
LeadA possible client who is not yet a contact, with a sourceContact, Property, Place, User
PropertyA physical parcel or homePlace
SaleA recorded sale of a propertyProperty, Place
PlaceA state, city or neighborhoodPlace
ListingOne effort to sell or buy a propertyProperty, Place, User
TransactionA buy or sell engagement, the thing people call a dealContact, Listing, Organization, User
TaskA unit of work someone owesContact, Transaction, User
ConversationAn email, text or call threadContact, Listing
Calendar eventA scheduled meeting or showingListing, Task, Transaction
SequenceA drip campaignContact
ListA saved group of contactsNone
DocumentA file in a transaction's vaultTransaction
NoteSomething a person wrote down about a contact or a transactionContact, Transaction, User
UserThe agent's accountNone
OrganizationThe brokerageNone

Your own stage, type, tag and source values show through. If your team calls a stage "Nurture" or tags people "Past client", an AI app sees those words, not ours.

What an outside AI app can read today

Four of the 16: contacts, transactions, tasks and conversations. You choose which of the four an app may read when you connect it, and you can change your mind later. The other types are used inside RealAnalytica. MLS listings, tax records and recorded sales are shared or licensed data, so they do not leave RealAnalytica through an outside AI app. Our guide on whether ChatGPT or Claude can use MLS data covers the rules.

Take "What closes this month, and what is still open?" Connected to an ontology, an AI app can:

  1. Ask which record types exist, and see that a transaction has a closing date and links to tasks.
  2. Find transactions that close this month.
  3. Follow the link from each transaction to its tasks and keep the open ones.
  4. Answer with the deals and the open work, read from the records.

No step needs a guess, and no step needs you to paste anything into the chat.

What an Ontology Does Not Do

  • It does not fill gaps. If a contact has no phone number or a deal has no closing date, the AI cannot find what was never recorded.
  • It does not decide access. The ontology says what exists. Permissions say who may see it. Ours come from the same checks the app runs when you sign in.
  • It does not replace judgment. An agent can find the three clients who went quiet. You still decide what to say to them.

What This Means for You

The Bottom Line

An AI model is only as useful as what it can reach. A real estate ontology is the plain list of what there is to reach: the record types, what each holds, and how they link. It turns a pile of exports into a business an AI agent can read. RealAnalytica starts with a 7-day free trial (card required), or you can book a demo.

Frequently Asked Questions

What is an ontology in real estate?

It is the list of record types a real estate business keeps, what each one holds, and how they link. Typical types are contacts, leads, properties, listings, transactions, tasks and conversations. A contact links to their deal, their tasks and their emails, texts and calls, so an AI agent can follow those links to answer a question.

What is a unified data layer for real estate?

It is one place where data from your CRM, inbox, phone and transaction software sits in one shape, so every screen and every AI tool reads the same records. The ontology is the part that defines that shape: which record types exist and how they link.

Is an ontology the same as a knowledge graph?

They are close. A knowledge graph is the records and links themselves, stored so you can walk from one record to the next. The ontology is the plan for it: which kinds of record and which kinds of link can exist.

How is an ontology different from the RESO Data Dictionary?

The RESO Data Dictionary is the industry standard for naming MLS data fields, built so MLS data can be exchanged between systems. A brokerage ontology also covers the daily work around a listing: your contacts and their stages, your deals, your tasks and your conversations.

Why do AI agents need an ontology?

An agent that knows only a list of actions can do things but cannot say what exists. With an ontology it can ask which record types there are, find the right record, and follow links to related records. It also keeps the tool list short: in RealAnalytica, six read tools cover every record type.

Does RealAnalytica's ontology include MLS data?

Listings, properties and sales are record types in the ontology, and Atlas works with them inside RealAnalytica. An outside AI app such as Claude or ChatGPT can read four types today: contacts, transactions, tasks and conversations. MLS, tax and sales data do not leave RealAnalytica through an outside app.

Can ChatGPT or Claude use RealAnalytica's ontology?

Yes, on any paid plan. Add https://www.realanalytica.com/mcp as a custom connector in the AI app, sign in to RealAnalytica, and choose what the app may read and do. Each record type is its own box, and every action starts switched off.

Share this article

One useful email a week

Practical ways to hand routine work to AI, from follow-up to closing. One short email a week.

Unsubscribe any time with one click.

Hire your AI workforce

Specialist Atlas agents handle the follow-up, the prospecting, and the routine work. You review, approve, and stay in control.

Card required; nothing is charged until the trial ends. Self-serve works where we carry your MLS. Not self-serve in your area yet? Book a demo and we'll set you up.