Analysis
Thomson Reuters launched a rebuilt version of its CoCounsel Legal product on August 20, 2026, describing it as a fully agentic AI experience meant to carry legal professionals from initial research through a finished, citable work product inside a single workflow, according to the launch announcement and wire coverage. The new version is built on Anthropic's Claude Agent SDK, and Thomson Reuters separately expanded its CoCounsel Legal integration with Claude via the Model Context Protocol this month -- deepening a partnership between the two companies rather than building the underlying model in-house.
- ## What's actually new
- The rebuild adds several features on top of the original CoCounsel, which launched in August 2025 and expanded to the UK in January 2026:
- Westlaw Brief Builder -- uses Westlaw Deep Research, KeyCite and Practical Law together to draft first-draft litigation briefs, proposing facts, arguments and supporting authority rather than only returning case citations
- Workspaces -- matter-specific document organization built for how litigation and transactional teams already structure case files
- Drafting Agent -- operates directly inside Microsoft Word rather than a separate web interface, meeting lawyers in the tool they already draft in
- Tabular Analysis -- reviews up to 10,000 documents against up to 100 questions in a single batch, aimed squarely at discovery and diligence review
## The competitive landscape Legal AI has become one of the more contested enterprise-AI verticals in 2026. Harvey, reported at roughly an $11 billion valuation as of March 2026, has built its business specifically around large law firms and built a head start in that segment. Legora has taken a similar large-firm approach. LexisNexis, Thomson Reuters's most direct legacy competitor, launched Lexis+ Protégé in February 2026 with more than 300 pre-built workflows and access to both GPT-5 and Claude models -- explicitly positioning itself as model-agnostic rather than tied to one lab, a contrast with Thomson Reuters's deeper, single-partner integration with Anthropic. Thomson Reuters's advantage against all three is distribution: CoCounsel ships bundled with Westlaw and Practical Law, products that are already the incumbent legal-research systems inside most large US law firms, giving Thomson Reuters a default-adoption path that AI-native challengers like Harvey and Legora have had to win deal-by-deal.
## What to watch Thomson Reuters has not disclosed pricing specific to this rebuilt version, and the rollout starting in the US only, with Canada, UK and Australia following later in 2026, means the near-term competitive test is domestic: whether Am Law 100 firms already paying for Westlaw treat the agentic rebuild as included value that deepens lock-in, or as a reason to evaluate a specialist competitor instead now that the baseline expectation for legal AI has moved from search-and-summarize to draft-a-brief. The real constraint on any of these products, including CoCounsel, remains the same one the entire legal-AI category has wrestled with since 2024: a first-draft brief with an unverified citation is a malpractice risk, not a productivity win, so the actual adoption curve depends on how much attorney review time these tools save net of the review they still require.
