If you are trying to evaluate AI tools for commercial real estate, the first question most people ask is the wrong one: will this save me time. Two years ago that was a fair filter. The general-purpose models already clear it now, which means any paid, CRE-specific tool has to earn its place on a harder test. After testing several hundred of these tools, I run every new one through the same four questions before I will pay for it. The questions go in order, and most tools fail on the first one.
Will it actually grow my business
Saving time used to be the whole pitch. A vendor would show you a task that took four hours and do it in ten minutes, and that was enough. That era is over. Claude, ChatGPT, and Gemini already compress most document and analysis work on their own, for the price of a basic subscription. So when a CRE tool leads with time savings, it is competing against something you can already do for twenty dollars a month.
The question I care about is whether the tool grows the business. Does it surface deal flow I would not have seen. Does it match a listing to the right buyer faster than my own memory does. Does it let me run three BOVs in the time one used to take, so I can chase three listings instead of one. Time saved is an input. The deals or revenue it produces are the only output that pays for a subscription. If a demo cannot draw a straight line from the feature to a deal or additional revenue, that is the answer.
Is there proprietary data underneath it
This is the question that separates a legit tool from a wrapper. A model is a commodity. Everyone has access to the same Claude and the same ChatGPT, which means any capability that lives entirely inside the model is something a competitor can replicate by Thursday.
What cannot be replicated is data. A tool sitting on a proprietary set of rent comps, sales comps, ownership records, or lease data has something the base model does not, and something a competitor cannot spin up overnight. That is the durable moat. When I evaluate a tool, I want to know what data it has that I cannot get from a general model, and whether that data is good enough to depend on. If the honest answer is that the tool is mostly prompt engineering on top of a public model, I can build that myself in an afternoon. Proprietary data also tends to be where these companies survive or die. The wrappers churn out fast and disappear. The data businesses tend to stay.
How easy is it to get on, and off
Most buyers only ask how hard a tool is to start using. The more important question is how hard it is to leave.
This market moves fast enough that a tool I commit to today may be second-best in a quarter. In AI right now, even a week is a long time. So I treat long contracts, heavy onboarding, and anything that locks my data inside a platform as real costs. If switching means I lose my history, my templates, and my data, the vendor has quietly made the decision for me to stay, whether or not a better option shows up. I want my data exportable and my exit cheap. A tool confident in its own product will let you walk. A tool that needs a twelve-month contract to hold you is telling you something about how it expects to compete.
What does the pricing actually look like
Subscription fatigue is real, and CRE professionals feel it. Add five AI tools at fifty to a few hundred dollars a month each and you are carrying a software bill that needs its own line in the budget, often for tools you touch twice a month.
Usage-based and credit pricing has started to win for a reason. Paying for what you use, instead of a flat monthly fee you mostly waste, matches how this work actually happens. Deals are lumpy. Some months you underwrite twenty, some months you underwrite two. A pricing model that flexes with that is worth more than a discount on a flat plan you will forget you are paying for. Subscriptions are fine when I use the tool. The problem is the recurring fee for the one I touch twice a month and forget about.
The filter behind the filter
All four questions sit on top of one habit, and it is the one most teams skip. You have to measure the tool against the work. Without a way to measure the result, you cannot manage the spend, and a lot of what passes for AI productivity right now is false productivity: activity that feels efficient and never shows up in a closed deal or an hour you can actually point to.
Before you buy, decide what the tool has to improve, and how you will know. More qualified deals this quarter. A BOV turnaround cut from a day to an hour. Pick something you can actually count. Then let the four questions tell you whether the tool earns a place in your stack, or whether you already own the only tool you need. That last part surprises people. For most of the work, the general model you already pay for is the tool. The specialized purchase only earns its keep when it clears all four.
If you want the tested workflows behind this, the exact tools that pass this filter for brokers, analysts, and developers, that is what we publish every week inside the AI for CRE Collective. Come find us.
