Thomson Reuters' CoCounsel deploys agentic AI for precedent-based agreement drafting and multi-document review, signalling a shift from AI search tools to AI workflow agents in legal practice
Thomson Reuters has published details of its March 2026 enhancements to CoCounsel Legal, its AI-powered legal research and drafting platform, introducing what it describes as 'agentic AI' capabilities grounded in deep legal expertise. The key new features include the ability to draft agreements based on a firm's own precedent documents and Practical Law content — meaning the AI can generate bespoke contractual drafts by synthesising a firm's internal templates with Thomson Reuters' curated legal content library — and enhanced multi-document review capabilities that allow legal teams to analyse large, complex document sets simultaneously rather than sequentially. The platform also introduces tools for professionals to 'manage their expertise' by embedding firm-specific knowledge into the AI's outputs. Thomson Reuters frames the development around three principles: agentic AI grounded in legal expertise, capabilities rooted in the user's own knowledge and workflows, and tools built to 'elevate' rather than replace legal professionals. The shift from AI as a search and retrieval tool to AI as a workflow agent — one that takes multi-step instructions and produces substantive outputs without constant human re-prompting — represents the most commercially significant development in legal AI since large language models entered the market. This is directly relevant to how City firms structure their associate and trainee workflows, particularly in transactional due diligence, document review, and agreement drafting.
Why this matters
Agentic AI capabilities in platforms like CoCounsel Legal shift the productivity frontier in legal practice from incremental time savings on discrete tasks to wholesale automation of multi-step workflows — agreement drafting from precedent, document review, and legal research synthesis being the three highest-value applications. For City firms, this accelerates the structural question already visible in training and recruitment: if agentic AI can draft a first-cut SPA schedule from a firm's own precedent bank, the economic case for employing large cohorts of junior associates on drafting tasks weakens further. The 'why now' driver is the combination of large language model maturity and Thomson Reuters' integration of its Practical Law content library — giving CoCounsel a legal knowledge base that generic AI tools lack. Firms that integrate these tools effectively will compress turnaround times on transactional work, creating competitive pricing pressure on hourly-rate mandates. The governance and liability question — who is responsible when agentic AI drafts a materially incorrect clause — remains unresolved and will generate its own advisory demand.
On the Ground
A trainee engaging with AI governance work at a City firm would draft AI governance policy sections covering acceptable use of agentic drafting tools, mark up data processing agreements with AI vendors to ensure client data is not used for model training, and prepare vendor due diligence questionnaires assessing CoCounsel and comparable platforms against the firm's information security and professional responsibility standards.
Interview prep
Soundbite
Agentic AI that drafts from your own precedents isn't a search tool — it's a junior lawyer that never sleeps and never bills.
Question you might get
“If a City firm uses an agentic AI tool to draft a clause in an SPA and that clause turns out to be materially incorrect, how would professional liability principles apply, and what contractual protections should the firm have in place with the AI vendor?”
Full answer
Thomson Reuters has upgraded CoCounsel Legal to deploy agentic AI — systems that execute multi-step workflows autonomously rather than just answering individual queries — including drafting agreements from a firm's own precedent bank and conducting simultaneous multi-document review. This matters because it moves legal AI from productivity enhancement to genuine workflow substitution on tasks that have historically been the domain of junior associates and trainees. For law firms, the commercial pressure is twofold: clients will expect faster turnaround and lower fees on commodity drafting tasks, while the firms themselves must invest in AI governance frameworks to manage quality, liability, and data security. The wider structural shift is the convergence of general AI capability with domain-specific legal content — Thomson Reuters' integration of Practical Law content into CoCounsel's outputs gives it a legal knowledge moat that generic ChatGPT-style tools lack. Firms that build AI governance infrastructure now will be better positioned to deploy these tools confidently, rather than reactively banning them after an error reaches a client.
Sources
My notes
saved