RAG vs LLM: which one does a property business actually need?

Most conversations about AI in property management start with the same question: should we just use ChatGPT, or do we need something built around our own data? The honest answer is that these are two different tools, and choosing between them is mostly a question about where the answer should come from.

What a plain LLM does

A large language model (LLM) is trained on a very large amount of public text. It is good at language: drafting a tenant notice, summarising a long email thread, rewording a policy so it is easier to read, or explaining a concept. It answers from what it learned in training and from what you paste into the conversation.

What it does not have is your business. It has never seen your tenancy contracts, your maintenance history, your owner agreements or your service charge rules. If you ask it about them, it may still produce a confident answer, and that answer can be wrong.

What RAG adds

RAG stands for retrieval-augmented generation. The idea is simple. Before the model answers, the system first retrieves the relevant pieces of your own documents, then hands them to the model and asks it to answer using only that material.

The result is an assistant that can say, for example, "Under clause 7 of this lease, the notice period is ..." and point to the clause. The model supplies the language. Your documents supply the facts.

A simple test for choosing

Ask one question: if the answer is wrong, does it matter that it did not come from our records?

• If not, a plain LLM is usually enough. Drafting, rewriting, brainstorming, summarising something you paste in, explaining general concepts.
• If yes, you need RAG, or something stricter. Questions about a specific contract, a specific unit, a specific owner, or your own procedures.

Where each one fits in property operations

Plain LLM: first drafts of communications, meeting notes, job descriptions, training material, translating or simplifying text.

RAG: a staff assistant that answers questions about policies and procedures, a tenant-facing bot that answers from your building rules and FAQs, a tool that helps a leasing or finance team find clauses across hundreds of contracts.

What RAG does not fix

RAG is not magic. It is only as good as the documents behind it. Out-of-date policies, scanned files nobody can read and three versions of the same contract produce confusing answers, just faster. Before you build anything, tidy the source material and decide who owns it.

It also does not remove the need for checking. Good systems show their sources, so a person can confirm the answer in seconds. If a tool gives answers without sources on something that matters, treat it with caution.

A sensible way to start

1. Pick one document set that people already ask questions about, such as your operating procedures or a standard lease template.
2. Pick one team that will use it and one person who owns the documents.
3. Require sources on every answer.
4. Collect the questions the tool could not answer. They tell you what is missing from your documents.

Start there, learn what your people actually ask, and only then decide whether to widen it.