Meet ROSS, the new bankruptcy robo-lawyer used by major firms
ROSS is barely three years old, doesn’t wear suits and hasn’t graduated from university with a law degree, but he — or really, it — has already been hired by several of the world’s largest law firms.
ROSS is an artificially intelligent computer system developed a few years ago by Andrew Arruda, Jimoh Ovbiagele, and Pargles Dall’Oglo at the University of Toronto as a plain-language legal research tool. ROSS was originally a submission in a global competition in which IBM challenged 10 universities to come up with commercial uses for its “Watson” artificial intelligence platform.
The U of T entry was a natural language legal research tool called ROSS — sort of like Siri on the iPhone, but better versed in legalese. ROSS isn’t an acronym, but rather just a name the students gave their creation. The students ultimately came second in the competition, but they knew that as a business idea, ROSS would be a winner. They created ROSS Intelligence Inc., secured funding from startup accelerator Y Combinator and set up shop in San Francisco.
The concept is catching on. ROSS is now being used by a range of law firms who pay monthly subscription fees to use the system. It’s used by solo practitioners who don’t have the time or resources to hire human research staff. For big firms, it provides a competitive edge by churning out research around the clock. Current users include mega firms Dentons, Latham & Watkins and BakerHostetler.
“Our goal is to have ROSS on the legal team of every lawyer in the world,” says Arruda, the company’s CEO. “Wherever you see that repeatable pattern-type of legal work, you’ll see machine learning and natural language processing.”
At its core, legal research is a repetitive task, and that’s just the sort of thing that’s tailor made for computerized automation. Each year, courts across the common law world churn out millions of legal decisions. Lawyers comb these cases looking for legal principles, doctrines and precedents that can help them resolve their clients’ legal problems.
In the old days, finding the right case was a labour-intensive process. Lawyers spent long hours sifting through case books in the library. The arrival of commercial legal databases in the 1980s and 1990s certainly sped things up by making it easier to find cases. But speed isn’t everything. To use these systems well, lawyers had to master cumbersome search codes. And even with cases in hand, lawyers still had to read through the voluminous output to find the nuggets of legal information they were looking for.
ROSS’s AI technology replaces the need for Boolean search terms and codes with ordinary, plain language. If a business lawyer needs to know the difference between the legal concepts of “loss” and “recoupment,” all the lawyer needs to do is ask ROSS, “What’s the difference between loss and recoupment?”
The system’s output goes beyond merely locating the names of helpful cases. It instead draws the lawyer’s attention to the key passages in the case that might best answer the lawyer’s question. The system also asks lawyers to “vote up” or “vote down” the search results so ROSS can learn how to better answer future questions.
“When you ask ROSS a question on the law, you ask it like you’d ask a human colleague,” Arruda says. ROSS breaks down the sentence to determine the intent of the question, then conducts the search. “Once the system understands what you’re asking, it starts to learn how to get better,” he says. “ROSS is dynamic, not static.”
The current version of the system is being used to help bankruptcy and insolvency lawyers working at U.S. firms. Future versions of the system will expand to cover other practice areas and jurisdictions, Arruda says.