Do You Really Need That Shiny New AI Tool?
AI is doing marvelous things in commercial real estate right now. But this story isn’t about that.
There are numerous examples to show that artificial intelligence is making inroads in commercial real estate.
For anyone following CRE’s adoption of high tech in the last decade, this may be hard to grasp. For a long time CRE was seen as a laggard in advanced technology adoption, an image that was finally shed years ago. AI, though, is the height of advanced technology and computing and more often associated with such cerebral applications as curing cancer or creating self-driving cars.
It is doing those things, with great fanfare in many cases, but the technology has a wide range of use cases, some of which are being applied in commercial real estate, ranging from the valuation of buildings, to running smart buildings to revenue management.
And indeed, a strong case can be made for their use. AI can make such granular decisions as analyzing the profitability of a potential office building in a designated market. Conversely, it can also help landlords better price apartment or housing units across portfolios. There are many, many examples of AI making a significant difference for a CRE firm.
But that is not what this article is about.
As AI becomes more commonplace in the industry, it would behoove potential users to ask themselves a fundamental question before embarking on a possible purchase: do you really need AI right now/? The truth is, AI is not for everyone and in many cases an alternative legacy application can do the job—perhaps not as well, but sufficient enough to solve a pain point.
“For many companies it comes down to that,” says says Alec Page, vice president of RET Ventures. “Oftentimes they will conclude that they don’t need cutting edge technology even with all that it can accomplish. Yes, they may be excited about what could be accomplished five years from now with AI, but these companies may find it more urgent to solve a problem this year.”
Companies considering an investment in AI should ask themselves a few basic questions before taking such an expensive plunge, says Michael Yurushkin
CTO and founder of Brouton Lab, which has worked with several commercial real estate companies. “If you need AI only to analyze the data rather than letting AI take actions on your behalf, you’re better off sticking to the older technologies, he says.
Also key is whether you have the right data in place.
“For example, AI can analyze existing commercial real estate market data and decide which properties to recommend for brokers to buy or sell based on the broker’s past sales and intent,” he says. This might well be a wise investment.
“On the other hand, if you are missing a sufficient amount of high-quality data, AI will be useless for your business. Without data, it will not be able to forecast risk or to perform high accuracy predictive analytics. In this case, you are better off with older technology and systems that you are already using rather than investing a hefty amount in AI.”
After all, real estate still operates largely with manual processes and manual data collection, says Comly Wilson, director of marketing for Enertiv.
“There is plenty of room to run with technologies that don’t need the advanced pattern recognition capabilities of AI, he says. Building safety, for instance, does not necessarily require AI, he says. “A combination of air quality sensors, HVAC sensors and people flow sensors can tell operators the moment air flow or particulates in the air begin to cross unsafe thresholds. This data can populate a tenant-facing dashboard so they’re confident that ownership is being proactive. All that can be done without the need for AI.”
Is it True AI?
Another, related issue to consider is whether an application is truly AI-based or rather a legacy application dressed up as artificial intelligence.
One way to tell if the vendor is marketing a true AI application is to ask them about algorithm details, Yurushkin says. “For example, how exactly the vendor implements a convolutional neural network to determine potential economic value for a property. Or what is the role of the feature engineering stage in attaining a high predictive performance?”
Another question would be what machine learning algorithms did the vendor test—machine learning algorithms include regression trees, k-nearest neighbors, neural networks among others—and what are the advantages and disadvantages of each of them, he continues. Perhaps most importantly, ask the vendor what data sets did it use to train and evaluate their AI, Yurushkin says.
You want to make sure the training dataset was big enough, Wilson says.
“AI is fundamentally pattern recognition that no human could hope to achieve. The best AI data scientist in the world couldn’t train an AI with a limited and/or unclean data set. A competent data scientist could run laps around him/her with access to a very large database.”
Also be sure to test out the application on your own, advises Shanka Jayasinha, CEO of EDGE AI. “Good salesmen will test it their own way to showcase the best performance of the application so what you need to do is throw some curveballs in there and see how the application reacts.”