Late into the evening on July 26, 3023, Thomson Reuters announced its $650 million acquisition of Casetext, a company that has recently made waves with CoCounsel, one of the first two legal domain-specific GPT-4-powered AI Assistants to burst onto the market earlier this year.
Casetext obtained early access to GPT-4, and was working to verticalize it for legal use cases six months prior to its public release. In the few months since it launched, CoCounsel has proved successful in capturing the early adopters of generative AI, selling into numerous law firms and corporate legal departments and with pilots under way in many others. Early estimates suggested CoCounsel had been beta-tested at over 40 legal organizations. Firms like DLAPiper, Fisher Phillips, Reed Smith, Skadden, Orrick, and Kirkland & Ellis have been openly touted as early users, alongside Microsoft and Ford Motors.
With firms scrambling to take some kind of action in response to the recent breakthroughs in AI, Casetext has seemed a safer option than foundational models like OpenAI’s GPT series, both because as a provider it has gained trust in the market and an ability to satisfy the rigorous security requirements of law firms, and because it is one of the few legaltech providers with access to a structured body of caselaw that makes generative AI useful in a legal context.
What will the acquisition mean in the context of a market that is already in more flux than we’ve perhaps ever seen?
Cutting Edge Technology with Deep Pockets for Development
Thomson Reuters has a history of acquiring significant players in the market, notably including Contract Express, HighQ, Practical Law, and ThoughtTrace. With innovation labs, one of the most extensive legal content collections, an international footprint and a large budget, Thomson Reuters has been intent on evolving beyond research to become synonymous with the future of law. Earlier this year, the company announced plans to invest $100 million annually in AI. In late May, it announced that over the course of the year it would incorporate generative AI across its suite of legaltech products through a partnership with Microsoft. The first concrete use case to come out of this collaboration is a drafting assistant leveraging Thomson Reuters’ Practical Law content through the Microsoft 365 CoPilot in Word, giving rise to what is touted to be reliable generative contract drafting (still in beta).
So to those who have had their ear to the ground, this development cannot come as a surprise. And even for skeptics, there are significant benefits to the acquisition. As someone close to the companies is quoted as saying, the acquisition is “good for customers because it combines world class content with cutting edge technology.”
The Casetext team is reported to be staying on in full, ensuring that talent like Pablo Arredondo and Jake Heller will continue at the helm of CoCounsel. Even with significant investment and a partnership with Microsoft, Thomson Reuters was lacking the state-of-the-art technology and expertise that Casetext brings with it, whereas the Casetext team was leading the way in legal AI even before recent developments, leveraging neural networks in products such as AllSearch (now folded into CoCounsel). As part of the acquisition, Thomson Reuters has made it clear that it will continue to invest deeply in the CoCounsel product, helping it to develop further while being committed to keeping it as a stand-alone product.
On the Casetext side, although the company has had access to sound legal content, this is not nearly as extensive or complete as the Thomson Reuters legal library.
Use of CoCounsel has thus far been restricted to US firms and companies, since the legal-specific content with which the model has been reinforced is limited by jurisdiction. For those in Canada, the United Kingdom, and Australia who have bemoaned the narrow availability of CoCounsel, the acquisition likely means earlier access to the complete solution than would otherwise have been possible. The introduction of Thomson Reuters content will allow for faster expansion across regions.
Improved Reach and Customer Support
Casetext, for all of its best intentions and early success, is still a substantially smaller company relative to Thomson Reuters. With the huge demand in the legal market for suitable generative AI applications, it would have been hard for the team to keep up (at the time of acquisition, Casetext had 104 employees and reported over 10,000 customers, from solo practitioners to 40 of the AmLaw 100 firms).
As part of the acquisition, Thomson Reuters will provide distribution support and resources that will allow for fully-supported roll-outs, comprehensive training programs, and a high level of customer service.
The full extent of Thomson Reuters’ know-how around roll-outs and implementations, customer connections, and proprietary content assets will be brought to bear in supporting the CoCounsel product.
Security and Accessibility
Although Casetext is a trusted name in the industry, Thomson Reuters likely carries even more weight.
Thomson Reuters is a known entity at almost all law firms and many legal departments. From a licensing perspective, pushing through an additional product from Thomson Reuters will almost certainly be easier for the various stakeholders involved in the decentralized procurement processes in law firms than it would be to introduce a product from a new vendor, especially given that products built on generative AI and large language models are viewed by information security teams as being higher risk than most other solutions.
Access to Justice
One of the enormous opportunities for the legal industry when it comes to advanced AI is its potential to close (or diminish) the justice gap. Access to justice has been called out as a key priority in the acquisition. The Casetext team will be partnering with Thomson Reuters’ TrustLaw entity, which is the pro bono arm of the Thomson Reuters Foundation (and the world’s largest pro bono network), with an eye to leveraging CoCounsel in support of legal aid across the United States. The weight that these combined companies are able to throw behind such an initiative will be a force multiplier for access to justice, something that simply would not have been feasible at this scale without the acquisition.
Although there will be naysayers who despair of another sale to one of the large research behemoths, it’s important to view this acquisition in the context of recent technology evolution.
The development of large language models, including layering on additional domain-specific data and further research, comes at a steep price. They require an enormous amount of computing power and energy to train (at considerable expense to the environment), and the complexities around their deployment in the legal industry are broad and significant. Would it have made sense for Thomson Reuters to develop its own, stand-alone LLM? Doing so would have slowed the pace of development for both Thomson Reuters and Casetext at a critical juncture.
This is a time when pooling resources to generate the best, safest models for use in legal makes sense, especially if it means their deployment is better supported and the technology can evolve at pace. This acquisition pairs what might be the best example thus far of a domain-specific LLM for the legal industry with one of the most reliable content sets for legal research. That combination should allay some of the trepidation that many firms have had in relation to the technology, and pave the way for much broader adoption, which will serve to benefit the way that law is practiced and delivered – for commercial organizations and consumers alike.
In spite of the deal being signed on Monday, Thomson Reuters announced in its press release that closing of the transaction is subject to specified regulatory approvals and customary closing conditions and is therefore anticipated to occur in the second half of 2023.