We have all bought into the promise that AI is coming to commercial real estate. Decisions are going to be made with orders of magnitude more sophisticated than they are today. We are going to look back on our valuation processes in a few years like doctors before the discovery of germs, like mathematicians before the discovery of calculus, right? Well, only if we let it.
In order for AI and machine learning technologies to function properly and be the change we all hope to see in our commercial real estate world, they need to be trained with data. Fortunately, data availability is at an all-time high. This proliferation has enabled a lucrative industry to blossom, with companies now marketing their incredible wealth of data, available to other organizations to enable greater insights into trends, customer behavior, future investments and of course, to provide training for artificial intelligence. All of this has led to data becoming the new commercial property battleground.
While AI and machine learning have both been embraced by sections of the commercial real estate industry like the design and management of buildings, there is comparably less adoption in the valuations and modeling side. Investment funds and acquisition teams have a compelling use case for AI to be part of the investment decision process. Accurately predicting, for example, CAGR percentage across multiple states, asset managers could deliver an enticing market differentiator to a progressive investor for a product that is usually considered a bit of a commodity. In a world where everyone has access to the same data sets, analytical methods are the only differentiator.
So, is AI really the holy grail? Possibly, but imagine this scenario. There is a big meeting at Investment House Inc. One of the partners pitches an idea to the board that ‘he’ had. He wants to invest in industrial units in an emerging city downtown location, citing enticing return over the next three years. Sounds logical but the meeting takes a not so surprising turn when a senior partner asks for the partner’s rationale for making this recommendation. “I’m not sure, the machine told me it would be a good idea,” is not going to be enough. This approach will never replace the one we have now.
The problem is that most AI systems available today rely on the “black box” neural network architecture. While this method is fast and accurate and can consume vast quantities of data at lightning speed, turning out answer after answer, there is one major flaw. The opaque nature of “black box AI” means that the dynamics of the decision-making process are difficult to unravel. You only know what the AI’s answer is, not how it arrived at that conclusion or why it’s making that recommendation. As with most investment decisions, property investments have relied on the guile, experience and “gut-feeling” of the partner, and it’s almost impossible to fully replace that instinct by a machine. But that doesn’t mean it’s not possible to get close.
For AI to be successful across the industry there will need to be a seismic shift away from what the AI is telling you to why it’s telling you. The why is crucial to ensure that the artificial intelligence is fused with human observation and experience. The commercial property industry needs to embrace understandable AI, a technology that is transparent, verifiable and interpretable. With an understandable AI, the technology works in harmony with humans, providing the why with every what. This shift could be seismic for the industry but will only come if “black box” AI makes way for understandable AI.
Now imagine the same scenario from earlier. There is a big meeting at Investment House Inc and one of the partners pitches the same idea to the board. The senior partner predictably asks for the rationale behind the recommendation and is met with the answer, “we noticed from our AI that similar industrial investments were lucrative with an incredibly high correlation.” Because this partner utilized understandable AI, they’re actually in a position to intelligently defend their recommendation, making their decision much more likely to sway the others involved in the process.
Anyone that has ever had a good math teacher knows the answer is only as important as showing your work. How you got to a decision, especially for decisions as big and irreversible as the ones made in the property industry, is just if not more important than the course of action itself. We try to fit buildings into a simple math equation made up of cashflows, depreciation and net present values but in reality, a property and how it is tied to the places and people around it is much more complicated than we are able to compute numerically. AI can help make our equations handle more complexity but maybe its most important duty is to explain that complexity to us along the way.
