This Toronto startup uses A.I. to help lawyers predict trial outcomes

By using machine learning to rapidly analyze case documents, Blue J Legal’s software helps lawyers spot problems before their day in court

Judge’s gavel with laptop

(Larry Washburn/Getty)

Benjamin Alarie, who is both the Osler Chair in Business Law at the University of Toronto and founder of Blue J Legal, says he wants to create the equivalent of a racing video game for the legal profession – an ultra-realistic simulation that can help lawyers and accountants get a good sense of what they might experience in a real courtroom.

Tax Foresight, the company’s first product in that vein, went on sale last month.

The subscription software uses artificial intelligence to scan legal documents, case files and decisions to create a mock judgement. Users can thus get a more accurate assessment of cases and a prediction of how a particular dispute might go.

It’s an improvement over existing methods, Alarie says. The rapidly multiplying number of cases and documents is outstripping human researchers’ own capabilities, so tax professionals are having trouble making accurate judgement predictions.

“We use shortcuts, like looking at leading cases,” he says. “The software should actually take into account all the information from the case law. Rather than looking at the tallest trees in the forest, we want to take into account all of the trees.”

Blue J Legal’s genesis began in 2014 with IBM’s Watson Challenge at the University of Toronto. Alarie, an associate dean at the time, was approached to help judge the competition, a contest between students and startups to create commercial implementations for the technology company’s super-computer AI.

The challenge was won by Ross Intelligence, a company that created software that could dig through and digest thousands of case files. Alarie was intrigued by the possibilities of applying AI to tax law, so he struck up a seminar with some students to begin working on a separate tool.

“I could have sat there and clung to my chalk at the law school for the next 30 years and run out my career, but this is going to happen whether we’re involved [or not],” he says. “So I got really excited about it.”

By 2015, they had a prototype ready, which they tested with Deloitte and PwC. Thomson Reuters heard about the pilot program and came calling. Before long, Blue J Legal—so named for the “j” that follows judges’ names in legal writing—had an exclusive distribution deal with the IT company.

blue j

So far, the 12-person startup has been privately funded. Alarie believes his company’s AI tool will speed up and encourage settlements between disputing parties by giving them a better idea of where their arguments stand.

“The parties will get a better mutual trust because they’ll be able to say, if this were to go to court, what would a court say?”

Judges, for their part, might have fewer cases in the long run but the ones they’ll continue to preside over might be tougher – they’ll be the ones that are too close for the software to call definitively.

As far as Blue J Legal software being a racing-like game simulation, Alarie concedes that legal resolution tools aren’t likely to find a mainstream audience, conceding: “It might not be a best seller.”