Anthropic launched plug-ins for its Claude Cowork AI agent on Friday, enabling automation across legal, sales, marketing and data analytics. Photo illustration by Jonathan Raa/NurPhotoJonathan Raa/Reuters
The pattern should be familiar by now: An apparent breakthrough occurs in artificial intelligence that promises to upend traditional industries, and the market panics, driving down share prices in legacy companies.
On Tuesday, data providers and legal software companies had their turn after Anthropic released productivity tools for lawyers. Many companies that provide databases, analytics or other tools for the legal industry took a hit in markets around the world, as investors grew more skeptical of software providers of all stripes in the face of AI. Two State Street SPDR exchange-traded funds tracking software, services and financial data stocks lost a total US$300.6-billion in market value after the news, The Wall Street Journal reported.
The sell-off is not unlike what happened last year when Chinese company DeepSeek released cheap, efficient AI models, hitting the share prices of companies tied to building resource-intensive data centres. The reaction turned out to be overblown.
Market Factors: Worried about asset bubbles? There’s an index for that
Whether that proves to be the case again is unclear because of the deep uncertainty around the winners and losers of AI. Hardware companies such as Celestica Inc., which builds equipment for AI data centres, have performed well over the past year because their products are in demand. Software developers are another story. Toronto’s Constellation Software Inc., which acquires companies that build applications for niche industries, has seen its stock price fall more than 50 per cent in the past 12 months owing to concerns that AI coding tools make it easier for competitors to win business, and for the company’s own customers to make custom tools.
The market reaction Tuesday also highlights a tension between the frontier AI developers such as Anthropic, OpenAI and Google that build general-purpose large language models (LLMs) that can complete a range of tasks, and the companies that build products on top of these very same models.
Anthropic, for example, recently released a plug-in for its generative AI platform that it said can handle legal tasks such as reviewing documents, flagging risks and tracking compliance. That represents a possible threat not only to well-established legal tech companies, but a host of AI startups that are building tools for lawyers.
Thomson Reuters Corp., which is investing heavily in AI, saw its share price fall 16 per cent on the Toronto Stock Exchange on Tuesday. (Woodbridge Co. Ltd., the controlling shareholder of Thomson Reuters, also owns The Globe and Mail.) CS Disco Inc., an AI-powered legal services company in the United States, dropped 12 per cent. Meanwhile, LexisNexis owner RELX and the Netherlands’ Wolters Kluwer, which provides legal analytics services, fell 15 per cent and 13 per cent respectively.
The question is what is stopping Anthropic and its ilk from going even deeper into the legal tech space – or any other sector, for that matter – and building custom software on top of their models, potentially putting their own customers out of business.
“These applications that are simply a wrapper around what already exists in an LLM, I do not understand what the enduring moat is,” said John Ruffolo, founder of Maverix Private Equity. “You are building your business on top of another business that could compete against you.” That’s part of the reason why Maverix has stayed away from backing AI startups, even as valuations have soared.
How AI is both hurting and helping the next generation of workers
But when it comes to law, there are reasons why AI developers may not make much of an effort any time soon, said Scott Stevenson, co-founder of AI legal software company Spellbook. The market may not be big enough to be worthwhile, for one thing, but the specialization that some companies provide is not so easily replicated. Spellbook, for example, boasts some 4,000 customers that use its AI tools for contract reviews.
“The model providers will start building some light tooling for different verticals, but what Anthropic does is very thin,” Mr. Stevenson said. “Spellbook is like a toaster. We do one thing, and we do it well.” With every new LLM that’s released, Spellbook’s business has only grown, as law firms become more interested in specialized tools, he said. But any advancement in general-purpose AI systems forces every other company to up its game.
Other sectors may be more vulnerable to disruption. Companies such as Cursor, Replit and Lovable have achieved massive valuations thanks to the popularity of their AI coding tools, which can build apps and websites based on plain-language instructions. But these products make use of LLMs from the frontier developers, while Anthropic’s Claude Code has become the de facto platform for many software developers. OpenAI has its own coding platform called Codex.
For these large developers, building coding tools may be more straightforward than contending with complicated integrations with other software or processes that exist in other industries, and more lucrative, too, given the size of the software market. Plus, Mr. Stevenson said, AI companies are heavy users of these tools, so there is an incentive to keep improving them. “I do think they will compete with Cursor and others for sure,” he said.
For investors, it’s not entirely clear which sectors are going to be most affected by AI, let alone which companies. The S&P North American Technology Software Index, for example, is down about 19 per cent so far this year. Shares in video-game companies such as Take-Two Interactive Software Inc. and Roblox Corp. tumbled in late January after Google released Project Genie, which can build digital worlds based on text prompts.
“The competition in AI is the most ruthless competition in the history of technology, maybe,” Mr. Stevenson said. “The only moat right now is speed.”