How AI Is Being Used To Read Leases, Other Real Estate Documents

Originally Appeared on GlobeSt.com by Erika Morphy

“A lot of people tend to give heavily-weighted promises about AI and where it is going.”

CHICAGO–A typical client for Nicholas Bartzen, an associate with Levenfeld Pearlstein’s Community Association Group, would be a condominium representative whose building has anywhere from four to 500 units and whose board has a question that needs to be responded to quickly. The answer can most likely be found within the condo’s governing documents but as Bartzen tells GlobeSt.com, “the way these documents have been written is anything but uniform.”

To find the clauses that he needs within these documents, Bartzen turns to an application called Diligen, which uses machine learning — a type of artificial intelligence that self-corrects and learns as it receives more information — to scan the governing document. To be sure there are other applications on the market that offer similar AI-driven scanning services; Bartzen likes Diligen because he says it is aimed at his niche practice of law. Other applications that he vetted, he says, “didn’t have the algorithms that I need for my clients.”

AI Makes Its Way Into Real Estate Tech

It has become a cliche to say that real estate has been slow to adopt technology. In the last few years, though, the industry has made strides in many areas — apps now proliferate in such areas as tenant-landlord communications or asset management.

Now there are signs that AI is starting to infiltrate bits and pieces of the commercial real estate tech as well. Steve Weikal, head of Industry Relations at MIT Center for Real Estate, which maintains a database of about 2,000 proptech startups on the market today, says he is seeing AI used in locational decision-making, in valuations and appraisals, in serving, in evaluating the risks of a particular loan — and as Diligen and other apps illustrate, the scanning of leases and other legal documents.

The word ‘scanning’ though is something of a misnomer here. All told Diligen uses about 100 different algorithms to scan a contract and then extract and identify the relevant provisions, Laura van Wyngaarden, co-founder and COO of Diligen, tells GlobeSt.com. “Our system is trained to recognize certain legal concepts.” At its most basic level, she says, the software is useful in any context in which people must review a lease.

For instance, a shopping center owner that is signing on a new tenant needs to know whether the tenant is proposing terms that may conflict with any agreements with existing tenants. “That owner would have to check all of the other leases very thoroughly and very quickly in order to be able to say yes or no to this tenant on its particular terms and negotiate a good deal,” van Wyngaarden says. Such due diligence can be very slow and painstaking when it is done manually, she adds.

Another example is CREModels, which is also using AI in the processing of leases into abstracts, says Managing Director Mike Harris. “We process lease data, taking it from the raw lease forms and turning it into lease abstract documents,” he tells GlobeSt.com.

Harris as well lauds the efficiency and time-saving benefits of AI used in this manner but he also notes that the system doesn’t always flag the right clauses — AI is not perfect, at least not yet, Harris says.

The Overhyping Of AI

Indeed, Harris takes pains to explain what is possible and what is not, with AI at the moment. And kudos to him for that — AI, in general, tends to be grossly overhyped by people who should know better, a point to which Harris alludes. “A lot of people tend to give heavily-weighted promises about AI and where it is going. Many people will paint a picture of something that may be possible in five years but they will have it sound like it will happen tomorrow.”

For instance, right now data extraction is a doable feature — but data abstraction (meaning you have taken the raw data you have extracted and turned it into a quality, human-readable summary document) can be difficult, he maintains.

For that reason, Harris has come to the conclusion that any practical application of AI works best in combination with humans. “Left to its own devices it can take you into some really odd places.”

nada