A grad student in a real estate program asked me recently whether he still needed to learn the fundamentals, given where the tools are going. He could already get an AI to spit out a pro forma. Why memorize how one works.
I have been thinking about that question for weeks, because the honest answer cuts against most of what the AI conversation in our industry sounds like right now. Here is what I told him. AI is a multiplier, and it multiplies whatever you bring to it. Hand it real skill and it produces something close to a superhuman version of your work. Hand it confusion and it produces a faster, more confident, better-formatted version of being wrong.
The multiplier runs both directions
Most of the excitement about AI in commercial real estate assumes the tool raises everyone to the same level. It does not work that way. The same model that helps a sharp underwriter pressure-test fifty scenarios in an afternoon will help a sloppy one generate fifty scenarios built on a broken assumption, and the second person will not catch it, because they never understood the assumption in the first place.
I have watched this happen. Someone who does not really grasp the difference between a gross and a net lease asks an AI to underwrite a deal, gets back a clean-looking model, and has no idea that the expense load is wrong. The output looks more authoritative than anything they could have built by hand. That is the trap. The tool did not make them more right, only more confident, and it dressed the error up well enough that nobody in the room questioned it.
Garbage in, garbage out is an old idea. AI just industrializes it. Layer AI on a shaky foundation and what you get is more polished garbage, produced fast enough to make it harder to catch.
Why fundamentals get more valuable, not less
The instinct is to assume that as the tools get better, the underlying knowledge matters less. In practice the opposite is happening.
When everyone can generate a model, the person who can tell whether the model is right becomes the valuable one. Judgment is the scarce input now. Knowing that an insurance number is too low for an old building in this market, that a rent assumption is fantasy, that the debt structure does not fit the business plan, those are the calls a model cannot make for you. It does not know your market the way you do, and it will not push back unless you have trained it to.
I tell people the same thing I would tell my own kids going into this business. There is no shortcut from AI. If you do not understand the fundamentals, AI will not save you; it will expose you, faster and more publicly than the old way ever did. A bad model used to stay on your own laptop. Now it gets emailed to an investor with a confidence it never earned.
Synthesizer, not replacement
The framing I keep coming back to is that AI should be a synthesizer, not a replacement. It is at its best when it takes work you already understand and extends it, runs the variations you would have run with ten more hours, organizes what you know into something you can act on. It is at its worst when it stands in for understanding you never built.
That distinction is the whole game for anyone early in a career. Use AI to do the fundamentals faster once you actually know them. Skipping the learning shows up later, in a deal that did not pencil the way the pretty model promised, and by then the cost is real money instead of a grade.
What to actually do about it
If you are early in your career, the move is to become genuinely good at the underlying work and to be the person at your firm who knows how to point these tools at it. Learn to underwrite by hand. Read the actual leases. Understand why a number is what it is before you let a model generate a hundred versions of it. None of that is glamorous, and all of it compounds.
Then go be the AI person at your firm: the one who can take a tool, aim it at a real problem, and judge whether the answer that comes back is any good. That combination, real fundamentals plus fluency with the tools, is rare right now, and it is going to be worth a great deal as the people around you lean harder on outputs they cannot check.
The operators getting the most out of AI today tend to share one trait. They did the unglamorous work first, learned the fundamentals cold, and only then handed the work to a machine that could finally multiply it.
Every week I break down what is actually working with AI in commercial real estate: the workflows, the tools worth paying for, and the ones that are not. If that is useful, the AI for CRE Collective is where we go deep on it.
