One of the most potentially transformative new technologies in the legal field was on display over the summer in a conference room at Milwaukee’s von Briesen & Roper. There, an executive from the small startup ROSS Intelligence led a demonstration of the artificial intelligence system that the company has created for legal research in the bankruptcy field.
Thomas Hamilton, vice president of strategic partnerships at ROSS, was participating in the June meeting remotely from the company’s offices in San Francisco. He typed a question into a search bar projected on a screen: “When can student loans be discharged in bankruptcy in Arizona?”
The system retrieved passages from several rulings that most nearly summarized their holdings. The results showed that courts use a three-part test that looks at whether the loan has become an undue hardship.
Randall Crocker, the CEO of 150-lawyer von Briesen & Roper, watched the demonstration with satisfaction. “This product operates more intuitively” than traditional legal research tools, he says. “It helps us provide a better legal product. We’re kind of excited about it.”
Crocker says he believes that ROSS will allow smaller firms like his to be more competitive, by offering more efficient services: “It’s just the rational next step for firms like ours that think they can compete globally.”
ROSS, which celebrated its one-year anniversary in June, combines the Watson artificial intelligence technology developed by International Business Machines Corp. with its own proprietary innovations, aiming to create a cheaper and faster research tool for lawyers. On its website, the company promotes the ROSS system as “Your Brand New Artificially Intelligent Lawyer.” The company maintains that its “machine learning” system, in which the computer becomes continually better at recognizing relevant case law and other resources, produces better results more quickly.
Von Briesen & Roper, which allowed a reporter to watch the demonstration, is one of a handful of law firms to license and test ROSS, and the firm has been using it for client matters for several months. CIO William Caraher says that he has witnessed ROSS’ performance improve over time. “It’s gone from a simple interface that produced interesting results to really on-point results,” he says.
Caraher met ROSS co-founder Andrew Arruda before ROSS was formally launched, when the young entrepreneur and his team won second place at the Watson University Competition with an early legal research application using Watson. Caraher was so impressed that he contacted Arruda: “I said, ‘Whatever you do, we want to be part of it.’”
To succeed, ROSS must compete against the traditional giants of the $8 billion-a-year U.S. legal research industry, such as Lexis and Thomson Reuters Corp., which owns Westlaw. Not only are those companies already deeply entrenched in the legal marketplace, but they’ve been integrating aspects of artificial intelligence into their products for years (although not using Watson), and have more powerful products on the horizon, according to Mike Dahn, global head of Westlaw Product Management. In 2010, he noted, the company released its WestSearch product, which allows users to pose questions in natural language; this February it launched two new features called folder analysis and research recommendations, which use machine learning systems to more quickly direct users to relevant results. “They dramatically shorten research time,” says Dahn.
For its next step, Thomson Reuters is using Watson technology to create a product for compliance officers that will analyze situations and spot issues, according to Thomson Reuters’ Eric Laughlin, who is heading the Watson project. For instance, he said, an executive working on a cross-border merger could identify regulatory issues that are raised by merging employee data. The company plans to release the product to a select group of companies and law firms this fall, and make it generally available in 2017. The company is also exploring the potential of Watson-based technology for legal research and other applications.
LexisNexis also says that it’s moving forward on AI-based search technologies. Jeff Reihl, chief technology officer of LexisNexis Legal & Professional, said in a statement that the company is investing in technologies such as machine learning, cognitive computing and natural language processing.
Caraher, the von Briesen CIO, says the advantage of ROSS is that it already uses the Watson platform, while its competitors so far don’t. “It has a smarter engine in understanding what you’re asking it,” he says.
Several Am Law 200 firms have been testing ROSS’ technology in the bankruptcy area, which is the first specialty targeted by the company, and at least two of them, Latham & Watkins and Baker & Hostetler, have also licensed ROSS for commercial use. ROSS CEO Arruda, a 27-year-old graduate of the University of Saskatchewan College of Law who worked briefly as a lawyer in Toronto, says the company is considering applications for tax, intellectual property, securities law, family law, labor and employment law, and insurance.
“Our goal is to have ROSS on the legal team of every lawyer in the world,” says Arruda.
Latham and Baker & Hostetler declined to give demonstrations of ROSS. When asked about it, Peter Knight, the head of Latham’s bankruptcy group, said he’d never heard of ROSS and that his bankruptcy lawyers weren’t aware of it. But he was intrigued. “That sounds interesting,” he says. “I’d like to learn more about it.” Baker & Hostetler declined to discuss ROSS.
During the demonstration at von Briesen & Roper, Crocker admitted that he hadn’t used ROSS himself, instead leaving the experimentation and research to younger lawyers. One of those lawyers is third-year associate Patrick Greeley, who participated in the demonstration and praised ROSS for its ease of use, noting that it allows him to ask questions using natural language. “In five minutes you can be trained,” says Greeley. (The associate has since left von Briesen.) The product has no user manual; the company trusts that users, especially tech-savvy younger lawyers, can figure it out themselves. ROSS also constantly monitors developments in the law to alert users if there’s been a change that’s relevant to a recent search.
Von Briesen CIO Caraher says that he believes ROSS is easier to use and produces better results than WestSearch’s natural language searches. “Our users mostly utilize keywords and Boolean logic-type queries for the best Westlaw research results,” he says. “This may speak to the effectiveness of the Westlaw natural language search.”
One of ROSS’ newest features has the potential to be the biggest game changer, offering the possibility of eliminating even more work now done by associates. “Where it really shines is with the memo,” says Greeley. If he types in the phrase “Write me a memo” in front of a search question, ROSS will deliver a memo on the topic by email.
These memos aren’t as polished as what an associate might produce, Greeley says. “If I gave this memo to a senior partner, he probably wouldn’t be happy with me,” he acknowledges. But he says that the memo feature was still hugely promising. “ROSS can make mistakes, as we make mistakes,” he says. “But this is a great tool for starting research.”
Lawyers can give ROSS’ memos a thumbs-up or thumbs-down, so that the system can learn and improve future searches. One memo produced at von Briesen & Roper appeared to give conflicting answers to a question about vendors who have an open order with a company and ship goods after the company files for bankruptcy. Near the top of the memo was a short answer that these vendors don’t get priority for payment; a longer answer later in the memo said that they can sometimes get priority. Greeley says he wasn’t concerned by these differing responses. “I get different answers for everything I research. You rarely find black-and-white answers in the law,” Greeley says. “Every time I’ve used the memo function, the correct answer has been in there.”
During the memo demonstration, ROSS’ Hamilton said that some human effort is involved as quality control, to double-check the memo’s information. In a later interview, Arruda notes that the memo function is still in an early stage.
“We still have a lot to do there,” he says. “We don’t talk too much about that feature. We want to focus on more mature features, like searches.”
Arruda declined to say much about ROSS’ price, except that it’s cheaper than traditional search products. “We work with firms to find a price that works with the ROSS features they need,” he says. Arruda says that his company has received millions in funding, but he didn’t want to reveal the amount. Funders include Y Combinator and Denton’s NextLaw Labs.
One person who closely follows technology developments in the legal world says he’s talked to people who have used ROSS. “They’re cautiously enthusiastic about it,” he says. “My guess is that they’re looking at this as something worthwhile to experiment with at a low cost, to keep an eye on the direction of the industry in the long term, rather than something that would be magnificently useful to them in the short term.”
They may also be waiting to see what ROSS’ competition comes up with. If Thomson Reuters and other large data companies get in the game, they could tap both their huge subscriber bases and the enormous research content they control.
Arruda claims these Goliaths don’t scare him. “Part of why you see us making progress is that we are nimble and agile,” he says. “When you’re a big company and you’ve built your product around old technology, and a new technology comes out, to change and adapt, you have to gut your underlying technology,” he says. “Let’s not forget that most companies in the legal research space started as publishing companies, not as technology companies. That gives us a great competitive edge.”
Contact Susan Beck at email@example.com.