Artificial intelligence in real estate is no longer just about asking ChatGPT to summarise a document.
We are entering a new phase — where AI doesn’t just respond to prompts, but securely processes sensitive data, orchestrates workflows, and even completes transactions autonomously.
In a recent webinar, AI expert Jason explored what this next phase looks like for real estate professionals. Here are the key insights.
1. Processing Large, Sensitive Data — Securely
Real estate professionals deal with:
- Rent rolls
- Valuation reports
- Due diligence documents
- Legal contracts
- Financial models
- Market comparable data
AI is exceptionally good at finding “the needle in the haystack” — identifying escalation clauses in 200-page leases, extracting key metrics from PDFs, or summarising portfolio performance across multiple spreadsheets.
But one critical question remains:
How do we use AI without exposing confidential data?
Uploading sensitive contracts into a public large language model (LLM) is often not acceptable from a compliance or risk perspective. This is where a technique called Retrieval Augmented Generation (RAG) becomes essential.
What Is RAG?
RAG is a method where:
- A question is asked.
- Relevant data is retrieved from your private documents.
- Only the relevant pieces are sent to the language model.
- The model generates an answer grounded in your data.
Instead of uploading entire datasets into an AI system, you “retrieve” only the context needed for that specific question.
This ensures:
- Greater accuracy
- Reduced hallucinations
- Better data control
- Lower exposure risk
This approach is already used in:
- Internal knowledge chatbots
- Legal document review tools
- Portfolio analysis systems
- Tools like Google’s NotebookLM
2. Public Models vs Self-Hosted Models: The Trade-Off
There are generally three approaches businesses are using today:
1️⃣ Direct API Use (e.g. ChatGPT API)
- Easy to implement
- Fast performance
- Contractual data protections
- Small residual compliance concerns for some institutions
2️⃣ Hybrid (RAG + API)
- Only relevant data sent to the model
- Improved security
- High performance
- Common in enterprise setups
3️⃣ Fully Self-Hosted Models
- Maximum data control
- No public model exposure
- Higher infrastructure costs
- Slower performance unless heavily resourced
One striking demonstration in the webinar showed the difference in speed:
- Cloud-based large models: ~3–5 seconds
- Local self-hosted model on a high-end laptop: ~39 seconds
The takeaway?
The computational power behind public LLMs is enormous. When you run models locally, you quickly realise how much infrastructure is required to replicate that speed.
Security and performance exist on a spectrum — and every organisation must decide where it sits.
3. AI Agents: Beyond Simple Prompt-Response
Most people think of AI as:
Ask a question → Get an answer.
But this is only “narrow AI.”
The next evolution is AI agents.
What Is an AI Agent?
An AI agent does not just answer questions.
It:
- Interprets goals
- Plans tasks
- Chooses tools
- Executes multi-step workflows
- Waits for dependencies
- Delivers outcomes
Think of it as an orchestration layer above large language models.
For example, an AI agent could:
- Analyse a property brochure
- Extract key financial metrics
- Pull comparable sales from external APIs
- Retrieve rental data
- Populate a financial model
- Calculate IRR
- Generate an investment summary
All automatically.
This moves AI from being a “smart assistant” to being a workflow engine.
4. Model Control Protocol (MCP): Connecting the Tools
To make agents powerful, they need structured access to tools.
That’s where Model Control Protocol (MCP) comes in.
MCP allows organisations to:
- Expose specific system functions (e.g. “search customer”, “create opportunity”)
- Connect CRMs, databases, APIs, or financial feeds
- Allow AI agents to safely trigger predefined actions
In simple terms:
- The LLM is the brain
- The agent gives it autonomy
- The tools are the hands
- MCP is the nervous system connecting everything
This architecture is particularly powerful for:
- Portfolio monitoring
- Risk detection
- Automated investor reporting
- Deal analysis workflows
5. Autonomous Commerce: AI That Transacts
One of the most forward-looking topics discussed was agentic commerce.
We are moving toward a world where:
- AI can search for products
- Compare options
- Make purchases
- Process payments
- Complete transactions
All without the user visiting a website.
Open standards like:
- Agentic Commerce Protocol
- Google’s UCP
are being developed to allow AI systems to transact directly with online stores via structured back-end access.
In the future:
- Your AI could book flights
- Reserve accommodation
- Purchase services
- Execute repeat orders
All autonomously.
For businesses, this means:
- Websites become part of a larger AI-driven commerce network
- Marketplace dependency may reduce
- Discovery may shift from search engines to AI agents
This has major implications for how property platforms, listing sites, and service providers operate.
6. What This Means for Real Estate Professionals
We are still early.
But the direction is clear.
In the near future, AI in real estate will likely:
- Automatically analyse deals
- Continuously monitor portfolios
- Flag risk signals
- Draft investor updates
- Retrieve compliance data
- Execute structured workflows
- Potentially transact property-related services
The competitive edge will not come from simply using ChatGPT.
It will come from understanding:
- Secure AI architecture
- Workflow orchestration
- Agent-based automation
- Data governance
Final Thoughts
AI is evolving from:
Tool → System → Autonomous Workflow Layer
Understanding concepts like RAG, AI agents, MCP, and agentic commerce allows real estate professionals to move from passive usage to strategic implementation.
The firms that learn to integrate AI securely and intelligently into their workflows will not just save time.
They will fundamentally change how deals are analysed, executed, and managed.
If you’re interested in learning how these tools apply specifically to real estate finance and investment workflows, this is exactly what we explore in our AI Applications for Real Estate course.
The next phase of AI isn’t coming.
It’s already here.
