Open any five real estate software websites this month and you will read the same sentence five times. AI-powered. AI-driven. Agentic. Your intelligent assistant. The homepages have converged so completely that the words have stopped carrying information — which is a problem if you’re the one deciding where to move your database.
This isn’t a complaint about marketing. It’s a practical shopping problem. When every option claims the same capability, the label can’t help you choose, so you need questions that the label can’t answer.
Here are four that work.
1. “What does it know about me that it didn’t know on day one?”
This is the question that splits the field, and almost nothing else matters as much.
There are two very different things being sold under one word. The first is a general model — the same one anyone can open in a browser — placed inside the software with a text box around it. It’s genuinely useful. It also knows nothing about your business. Every time you want output that sounds like you and references the right history, you supply that yourself, by hand, in the prompt.
The second is a system that has actually been given your material: how you write, who your people are, what happened with each of them, what you did last time in a situation like this one. The output difference isn’t subtle, and neither is the work difference. One requires you to reconstruct context before every task. The other already has it.
Ask for a specific answer. “What, concretely, has it been trained on?” If the answer is a version of “it’s very smart,” you’re looking at option one with better packaging.
2. “Show me what it does without me asking.”
“AI-powered” often turns out to mean there’s a chat window. You go to it, you type, it responds, you copy the result somewhere useful. That’s a tool, and a tool is only as good as your memory to use it.
The more meaningful question is what happens when you don’t open the software at all. Does anything get noticed? Does the system surface that a lead went quiet, that a file is missing a document, that someone you should call has been sitting untouched for six weeks? Work that only happens when you remember to trigger it is work that mostly doesn’t happen.
Ask them to walk you through a Tuesday where the user never types anything.
3. “Who owns the data, and what leaves the building?”
Your database is the asset. Everything else is software you can swap.
Two separate things to nail down. First: if you leave, what comes with you — a clean export of contacts, history, notes, and documents, or a CSV of names and emails that strips out the part that made it valuable? Second: where does your client information go when the AI processes it, and is it being used to improve a model that other people will benefit from?
These questions have real answers and vendors can give them. A vague answer is itself an answer.
4. “Who sets it up — and what does that cost?”
This one gets skipped and it’s the one that kills most rollouts. Powerful software that arrives as an empty shell with a help-center link gets configured for about a week. Then the old habits come back and you’re paying monthly for a database with an unused AI stapled to it.
Ask exactly who does the initial setup, whether a real person builds your routing and follow-up rules with you, what it costs, and what happens in week three when something needs adjusting. A “concierge onboarding” that’s a 20-minute screen share is not setup.
And while you’re at it: ask for the price in a number. If pricing requires a discovery call, the price is “whatever they think you’ll pay.”
The shortcut
If you only have time for one, use the first. Everything downstream follows from whether the system has been given your business or is merely sitting next to it.
The reason we’re blunt about this is that Theia Vault isn’t a CRM with AI features added — the whole design premise is a private AI trained on your material, with your data staying yours. That’s a claim you should stress-test with the four questions above, on us and on everyone else. We wrote up how we stack against the major platforms, with sources and dates, on the comparison pages. Read them skeptically. That’s the point.