P2P

37 I L T A N E T . O R G A rtificial Intelligence ("AI") seems to be everywhere these days. We see AI helping to correct our text messages and power suggestions on which products we should buy online or which new television programs we should binge. Those of us in the legal innovation community see articles everyday which discuss AI and how it will revolutionize the future of the legal industry. The purpose of this article is to discuss a recent project where I trained AI extraction technolo ("AI Software"), which many law firms license today, and deployed it to assist on a market study we conduct at Sidley Austin LLP 1 . Ideal Use Cases for AI Software: The best use cases for AI Software involve projects designed to expedite review of repetitive items. The typical use case that most law firms have found for AI Software is to help expedite legal due diligence. While there are a number of AI Software tools which specialize in extraction, the AI Software I worked with on this project is KIRA, which is trained to recognize over 1000 unique clauses and concepts across a variety of contract types, but can also be trained to recognize additional clauses. A few years ago, I thought that some of the market studies done in the Sidley real estate department would be a great potential use case to train AI Software. Sidley's real estate department conducts a series of "What's Market" studies which do a "deep dive" into the some of the most heavily negotiated clauses in Purchase and Sale Agreements, Joint Venture Agreements and Non- Recourse Carveout Guaranties ("Guaranties"). The market presentation for Guaranties focused on (i) the frequency with which the Guaranties contained certain acts which triggered liability for a guarantor ("Bad Boy Acts") and (ii) classifying each Bad Boy Act depending upon the level of liability it triggered for the guarantor. This Guaranty market study turned out to be a very good project to train AI Software for a number of reasons: • We had already established the list of Bad Boy Acts that existed in these Guaranties; • The Bad Boy Acts, regardless of lender, contained relatively standard language; • We had already collected market data on over 200 Guaranties, so we knew which Guaranties contained specific Bad Boy Acts; and • It was essential to partners in the real estate group that we keep this market study updated in an efficient manner since it would be non-billable work for our 1st year associates going forward Training AI Software: Thoughts on person training the AI Software: • The person training AI Software needs to be someone who really understands the clauses that the AI Software is being trained to recognize. I was a real estate lawyer for 10 years and had also analyzed hundreds of examples of these Bad Boy Acts to prepare the market study over the years. 1. There are many AI solutions for law firms today. This article focuses on AI extraction technolo typically used in legal due diligence. It is not going to focus on AI used for research, predictive analytics or natural language processing to improve searching a firm's document management system.

• Cover