Skip to main content
Open this photo in gallery:

Sarit Pandya, Partner, Real Property and Planning with the law firm McCarthy Tétrault, demonstrates the firm’s use of AI programs, on Feb 22.Fred Lum/The Globe and Mail

Commercial real estate lawyer Sarit Pandya needed some grunt work done. The partner at McCarthy Tétrault was acting for a lender who had financed a property deal involving two owners, and he needed someone to go through the co-ownership agreement, summarize the document and prepare a breakdown of the liquidity provisions.

Typically, this monotonous task would fall to a junior associate. But not anymore.

Late last year, McCarthy launched pilot projects for three new generative artificial intelligence programs within the firm. So instead of tying up an associate’s afternoon, Mr. Pandya was able to upload the relevant document himself, bang out a few sentences of instruction and get back exactly what he needed in seconds.

“I asked it a pretty vague and technical question and it got it right,” he said. “This is something I still might ask a junior to do as a learning piece. But when you’re in a rush, this was really easy.”

Few topics in the legal world are as buzzy right now as AI. Depending on the area of law, AI is either the bringer of doom – commoditizing work such as contracts, incorporations and wills – or a welcome innovation that will improve efficiency within firms. What everyone agrees upon is that AI, particularly generative AI, will be a great disrupter for the legal sector.

And for their part, McCarthy has decided to lean in. It began heavily investing in AI about seven years ago when it acquired Wortzmans, a niche law firm that specialized in e-discovery and data governance. The firm was rebranded as MT>3 and now operates as a division within McCarthy. MT>3 uses machine learning to build AI models that can analyze vast amounts of records and evidence, pulling out and analyzing the most relevant material.

“The traditional model for document review would have been an army of reviewers sitting in a room together, going through hundreds of thousands of documents. It could take months and be very expensive,” said David Cohen, a senior director with the firm’s innovation group.

“What we found is we saved time, we saved money and there was no diminishment in quality of the review.”

Now, generative AI – which can innovate and be creative – is the new frontier, according to Mr. Cohen.

In December, after several months of testing the tools with smaller user groups, McCarthy launched their generative AI programs across every practice group, region and staff level.

The first, ChatMT, is the firm’s own version of ChatGPT, but protected within McCarthy’s security network to keep client information secure.

The second is Microsoft’s Copilot system, which lawyers can use to do things such as: create a timeline of discussions with a client based on their e-mail history, generate a PowerPoint deck using uploaded files or prepare a summary of a Teams meeting.

The third is CoCounsel, which is what Mr. Pandya used for his contract analysis. The program’s advantage, Mr. Cohen said, is that last year, its parent company, Casetext, was acquired by Thomson Reuters, which operates a suite of law research platforms. This enables CoCounsel to pull from services such as Westlaw. (Both Thomson Reuters and The Globe and Mail are owned by Woodbridge Co. Ltd.)

Jake Heller is the co-founder of Casetext. He said that, at a high-level, CoCounsel is supposed to function as an AI legal assistant.

Mr. Heller himself is a trained lawyer. After graduating from Stanford Law School, he spent a year clerking for a judge at an appeals court and then worked as an associate at a large law firm, Ropes & Gray LLP. But having grown up in the San Francisco Bay Area, he had always been really into computers and coding. The more he worked in law, the more he realized how behind the legal profession was.

“The feeling I felt was that the tech was working against me,” he said. “I knew I could do something really trivial – find an open vegetarian Thai restaurant near my house – using Google. But if I wanted to find the one case that could exonerate my client. I’d be up until 4 a.m.”

This was the thinking behind Casetext, which Mr. Heller founded in 2013.

Two months into McCarthy’s pilot, Mr. Pandya says he’s using CoCounsel most often to speed-up the writing process.

“I can ask CoCounsel: ‘Draft me a letter to person X about topic Y with a relatively forceful tone,’ and it will produce something I can use as a starting point,” he said.

Mr. Cohen said the firm’s AI policy clearly notes that all output from AI software must be verified, “which is something that they would be doing if they were asking an articling student to prepare a research memo.” McCarthy is also selective about what AI tools it uses and for what.

“One of the reasons we’re comfortable with CoCounsel is that it’s not prone to hallucinations,” he said.

This is particularly relevant given a recent case out of British Columbia in which Vancouver lawyer Chong Ke was found to have used ChatGPT to prepare materials in a family law case. The document cited cases that didn’t exist. Ms. Ke said she was not aware that AI could generate fictitious case authorities – or as court heard, an AI hallucination – and that she was “deeply embarrassed” about the error.

Robin MacAulay, McCarthy’s chief professional resources officer, said that, with AI taking on a larger role in the day-to-day, the firm is working on new training for junior lawyers that will replicate some of the learning that comes with executing those seemingly mundane research tasks.

The upside, she said, is that their associates and students will be freed up to tackle more complex work earlier on in their careers – which is music to Mr. Heller’s ears.

“The legal profession used to be more of an apprenticeship. At a law firm, you’re taken on to be the future of the law firm. Over the last 50 or so years, we’ve gotten away from that,” he said.

“The point of associates wasn’t to grind them through a bunch of menial tasks until they quit or become partner, it was to train them. I think with AI, we’re going to return to that.”

Follow related authors and topics

Authors and topics you follow will be added to your personal news feed in Following.

Interact with The Globe