Why This Vendor is Avoiding AI

Every time there is a new type of technological marvel, the hype machine ramps up. Vendors, consultants, and analysts make predictions of how big the new thing will get and the need for companies to buy now and not miss the wave of success that’s bound to happen.

It’s always a chore to maintain balance. How well will the software work, especially with existing systems? Is it steady and reliable or in a testing stage? What should you do with it?

Generative artificial intelligence has shown some big issues, such as legal questions about using data, hallucination (making things up), and some mediocre results. But there is an additional problem that always happens with the new shiny toy: remembering that other tools might be at least as good and possibly even better in given applications.

John Cona, chief executive officer of software vendor F9Analytics, which provides predictive analytics for CRE price management, recently spoke with GlobeSt.com to discuss his company and how it works. Artificial intelligence isn’t one of the tools they use because, he says, there are better ones, and he claims that they built their own.

As a Microsoft partner, they see a lot of AI-type of software to test. He remembers sending a million records into a machine learning pricing system. “I didn’t get any metrics,” he says, so had no way to see what was happening or determine “how far off the results are. If a valuation is 30% off every five tries, it’s not a solution.”

Rather than what some large companies in the business do, which he characterizes as “gathering information from competitors, then averaging it and adding 3%.” Instead, the company uses its own mathematical algorithms incorporating long-standing techniques as well as their own intellectual property and models financial behavior of client properties and portfolios against detailed government data on historical apartment rents.

Cona and his firm also optimize for a different result than some of the other systems. Rather than looking for the optimum rent, they look for the optimum profit. “If you only look at the top line, you build class A buildings,” he says. Because you want to get the highest rents. However, what does that extra rent cost? There are amenities, higher prices for materials, expected services, and so forth. “Your rent has to reflect that,” he adds. “If it doesn’t, you’re underpricing.”

Could an AI program do this? Perhaps, depending on its structure. “It’s the difference between plausibly generating and what we’re doing is logically identify,” he says. “That’s why AI can’t do math.”

And sometimes you absolutely need something to do the math.