Thomson Reuters revenue rises 10% as use of AI-based software tools ramps up

Thomson Reuters Corp. TRI-T reported a 10-per-cent increase in third-quarter revenue and announced a partnership with a major U.S. law firm to develop proprietary artificial intelligence tools as customers ramp up their reliance on the company’s newer, AI-powered software.

The professional services and news provider brought in US$1.83-billion in revenue in the quarter that ended Sept. 30, with revenue from recurring contracts up 9 per cent.

Thomson Reuters also adjusted its financial outlook for the year based on the sale of a majority stake in its global print business to KKR & Co. Inc., which closed on Oct. 1, after the third quarter ended. The company now expects revenue to rise by about 9.5 per cent this year, compared to a previous 8-per-cent forecast.

Chief executive officer Steve Hasker said in a Tuesday interview that customers’ use of the company’s latest CoCounsel Legal software, which includes AI agents that can carry out complex tasks on their own, is “off to a very strong start” based on engagement metrics the company tracks “obsessively.”

The product’s active user base was up 60 per cent month over month in September and those users are completing 2.4 times the amount of work using CoCounsel, when compared with peak use of the product’s previous version.

“The intensity and duration of use has moved to an entirely different plane,” he said.

Startups in the legal software space and new tools released by leading AI companies such as Anthropic have intensified the competition for customers in the legal and corporate sectors, and put pressure on incumbents. But Thomson Reuters is making the case to clients that its products, trained on a trove of proprietary data, will be more accurate and trustworthy.

“The results continue to show strong demand for Thomson Reuters’ products, including Westlaw and CoCounsel, with no evidence to date of a slowdown from AI-native entrants,” Scotia Capital Inc. analyst Maher Yaghi said in a note to clients.

Thomson Reuters earned US$339-million, or 78 US cents a share, in the quarter that ended Sept. 30. That compared with US$423-million, or 94 US cents a share, in the same period last year.

Adjusted for certain items, Thomson Reuters said it earned 81 US cents a share, just shy of the 83 US cents analysts expected.

Woodbridge Co. Ltd., the Thomson family holding company and controlling shareholder of Thomson Reuters, also owns The Globe and Mail.

RBC Dominion Securities Inc. analyst Drew McReynolds said in a note to clients that the company’s quarterly results should be “largely neutral for the shares at current levels.”

But the company’s shares fell 7.4 per cent to $157.83 by midday on the Toronto Stock Exchange.

Thomson Reuters is betting that law firms – its largest segment of customers, along with corporate, tax and accounting clients – are in the early stages of a fundamental shift that will reshape how they spend their resources, Mr. Hasker said. Over the next five to 10 years, he expects firms will spend more money on technology, AI and automation as a share of overall revenue, and less on staffing and real estate.

“That’s a fundamental retooling of a law firm P&L,” he said, referring to the profit and loss statement that outlines a company’s mix of revenue and costs. And it could expand the size of the market Thomson Reuters serves.

In August, Thomson Reuters deployed its own artificial-intelligence model called Thomson LLM, which is tailored to legal professionals and was built in-house on a modest US$40-million budget. The company is looking to license the model to large law firms to use on their own systems, helping them keep control over their intellectual property and data.

On Tuesday, Mr. Hasker announced a partnership with Skadden, Arps, Slate, Meagher & Flom LLP, a leading U.S. law firm, to develop new applications on top of the Thomson LLM that Skadden would control.

The agreement is a first step toward using Thomson Reuters’ own AI software products and data to build “sovereign AI solutions” for customers, Mr. Hasker said.

The company expects to rely increasingly on its in-house model, which is cheaper to operate than many leading models built by Anthropic or OpenAI, to keep a lid on computing costs, allowing Thomson Reuters to “stick with all-you-can-eat pricing in most instances, versus usage-based pricing” for its products, Mr. Hasker said.

Pricing that is based on how much computing power users consume can be “problematic for private partnerships” such as law firms, he said.