Google Cloud launched Gemini Enterprise for Legal and Gemini Enterprise for Financial Services, agentic AI solutions built atop its existing Gemini models rather than a purpose-trained legal model. This came a day after Thomson Reuters unveiled Thomson, a proprietary model trained on decades of Westlaw, Practical Law, Checkpoint, and Reuters content, which reportedly beat Gemini 3.1 Pro, Claude Opus 4.8, and GPT-5.5 on some benchmarks. Google's offering centers on specialized agent skills, MCP integrations with tools like DocuSign, a partner network including Accenture and Deloitte, and governance controls, while Thomson Reuters pairs its own model with a multi-model routing strategy inside CoCounsel Legal, built on Anthropic's Claude Agent SDK. The two launches illustrate diverging strategies for specializing enterprise AI: differentiating above the model layer versus owning a proprietary, domain-trained model.
Table of contents
Specialized AI doesn’t have to mean a specialized modelOwning a model doesn’t necessarily mean going all in on itModel selection is now only one piece of the puzzleQuestions this post answers
What is Google Gemini Enterprise for Legal and what can it do?
Gemini Enterprise for Legal is an agentic AI solution from Google Cloud built on top of existing Gemini models rather than a new specialized model. It automates data discovery and Data Subject Access Request responses, tracks legislative updates and court dockets, drafts and updates policy and contracting playbooks, speeds contract review and negotiation, drafts NDAs, prepares court filings, and redacts documents. It uses purpose-built skills, secure MCP integrations with tools like DocuSign, a partner network including Accenture and Deloitte, and a governance control plane. Teams weighing agentic layers versus specialized models for legal workflows can track these shifts on daily.dev.
How does Thomson Reuters' Thomson model differ from Google's Gemini Enterprise for Legal?
Thomson is a proprietary model Thomson Reuters spent $40 million developing by further training an existing open-source foundation model on decades of its own proprietary content, including Westlaw, Practical Law, Checkpoint, and Reuters, plus expert evaluation. Gemini Enterprise for Legal, by contrast, is an agentic system layered on Google's existing Gemini models using skills, MCP integrations, and governance rather than a newly trained model. Thomson reportedly outperformed Gemini 3.1 Pro, Claude Opus 4.8, and GPT-5.5 on some benchmarks. Developers deciding between building on a general model or a specialized one can follow this debate on daily.dev.
What is CoCounsel Legal and how does it use the Thomson model?
CoCounsel Legal is Thomson Reuters' AI assistant for legal research, analysis, and drafting, built on Anthropic's Claude Agent SDK. Rather than relying solely on the proprietary Thomson model, it routes tasks across multiple models, letting Thomson lead when it has a real advantage, such as within the Tabular Analysis document-review product, and deferring to other models when better suited for a given task. Anyone architecting multi-model routing for domain-specific agents can follow related coverage on daily.dev.