Something changed in the last month. Firms that spent the past two years buying AI tools and pushing everyone to use them have started asking a harder question: is any of this actually working. The question is no longer whether it feels modern; it is whether the hours and dollars going into AI come back as deals, money, time, or margin. The answer, more often than people expect, is no. The culprit has a name: false productivity.
The agent who worked hard for two weeks and closed nothing
An investor-agent I spoke with, someone who runs a brokerage across a few Northeast states, told me about a producer on his team who came to him frustrated. “I’ve worked so hard the last two weeks, and I’ve gotten no deals,” the agent said. So he asked to see the work. The producer had spent two weeks inside Claude and ChatGPT, building systems, tuning prompts, formatting outputs. Then the realization landed: in two weeks he had not spoken to a single buyer or a single seller. He had been optimizing a system to avoid doing sales. As the broker put it, AI is an easy way to find the laziest path out of the work, and an even easier way to disappear down a rabbit hole of tool-tinkering that produces nothing.
What false productivity actually is
False productivity is the feeling of efficiency without the result. You are busy. You are using impressive tools. The work looks like progress. But you are creating more work, or you are doing work that was never going to move the number. I see it constantly: people convinced a tool is saving them time or making them money when the honest accounting says it is doing neither, and quietly pulling their attention off the things that do. AI is unusually good at manufacturing this feeling, because it has no brakes. It will go in whatever direction you point it, enthusiastically, for as long as you let it.
Why you can’t manage what you never measured
The reason false productivity survives is that almost nobody baselines the task. Take lease abstraction. Before you decide AI saved you time, you need two honest numbers: how long the abstract took you by hand, and how long it took with AI, including the human review and the time spent correcting what the model got wrong on the first pass. That second number is the one people skip. They count the AI minutes and ignore the cleanup. Without a baseline, you have a feeling, not a result. You cannot manage what you cannot measure, and a lot of AI spend is being managed on vibes.
The “AI everything” trap
The biggest version of this mistake is ambition pointed in the wrong direction. People try to AI everything. They want to build an agent that replaces the whole business, automate the entire pipeline, shoot for the moon on day one. That is where months disappear with nothing to show. The real returns almost always show up in the boring, less glamorous work: the repetitive document task, the follow-up reminder, the first draft of a proposal. Narrow and dull beats broad and ambitious, because narrow and dull is measurable and shippable this week.
The technology is real, which is what makes the trap dangerous
The point here works against the easy anti-AI read, because the wins are genuine. The same investor-agent gave one of his veteran brokers a skill that produced a 15-page offering memorandum in about five minutes, work that used to take an assistant two hours. He fed a 365-page zoning matrix and a town’s denial into Claude and got back the exact exception, on page 265, plus a precedent ordinance from the town’s own records, and got the denial reversed. Those are real hours and real outcomes. That is exactly why false productivity is hard to spot: it sits right next to the real thing, using the same tools, wearing the same look of effort. The only reliable way to tell them apart is the number.
What the disciplined teams are doing
The firms getting real returns share a habit. They set KPIs before they spend, baseline the workflows they care about, and check whether the investment shows up in the results. I had a version of this conversation recently with the CEO of a national brokerage, and even at that scale the honest position was uncertainty: people across the firm are going to self-serve these tools one way or another, and the open problem is measuring whether the output justifies the time and money. That shift, from assuming AI helps to proving it does, is the healthiest thing happening in CRE AI right now, and it started showing up in conversations only in the last month or so.
Run the audit yourself
If you want to check your own stack, it is not complicated. Pick one workflow and time it by hand. Then time it with AI, counting every minute you spend reviewing and correcting the output. If that number isn’t lower than the old way, the tool is costing you time, whatever it feels like. Most teams never run this, which is exactly why the false productivity keeps compounding.
