The latest legal AI news points in the same direction: firms want AI that is controlled, specialized, and tied to real work. Thomson Reuters' Thomson 1.0, LexisNexis' agentic capabilities, and Draftwise's legal ontology all show a shift away from generic assistants and toward systems built around firm data, matter context, and workflow execution.
The move from generic AI to specialized firm workflows
Thomson Reuters has launched Thomson 1.0, its own open weights LLM trained on its own data. The company says it is using the model first inside Tabular Analysis in CoCounsel Legal, with plans to extend it across its legal and tax portfolio.
For law firms and accounting firms, the strategic takeaway is not that every firm needs its own model. It is that AI becomes more useful when it is grounded in the work the firm actually does, rather than in a broad chatbot experience that does not understand the firm's data or processes.
Why control, context, and data structure matter
Thomson Reuters framed Thomson 1.0 as a more efficient path: start with a strong foundation, specialize it deeply, and keep it under your control. That same logic applies to custom AI and automation projects inside professional-services firms.
Draftwise's new Legal Ontology highlights a related point. The value is in connecting information across documents so lawyers can see negotiated positions, rights, obligations, and how one change may affect other relationships. That is the kind of structure firms need if they want AI to support real judgment, not just summarize text.
Agentic workflows are becoming the next interface layer
LexisNexis is rolling out agentic capabilities around its Legal Intelligence Engine, with a goal of letting users describe what they want to accomplish while the system applies models, agents, skills, and selected sources behind the scenes. It also aims to keep context with the matter and reduce jumping between systems.
For firm leaders, this is a practical signal that the next wave of AI may be less about chat and more about orchestrated workflows. That could include drafting, analysis, matter intake, research, and document production tied to approved sources and existing systems.
What this means for custom AI planning in professional services
The common thread across these announcements is not hype about bigger models. It is a focus on specialization, knowledge structure, and workflow design. Firms that want better returns from AI should think in terms of a few high-value use cases, clean source data, and clear boundaries around what the system is allowed to do.
For law and accounting practices, that means prioritizing the work where context matters most: intake, contract review, tax and legal analysis, document production, and knowledge retrieval. The firms that win will likely be the ones that build AI around process, not around novelty.
- Start with one workflow where context and consistency matter more than general conversation.
- Treat data structure and knowledge management as part of the AI strategy, not a separate project.
- Look for AI tools that can preserve matter context and work across approved sources.
- Evaluate vendors on control, specialization, and workflow fit, not just model size.
Sources watched
- TR CEO Steve Hasker on Thomson 1.0 (Artificial Lawyer)
- TR Launches Thomson 1.0 – Its Own LLM (Artificial Lawyer)
- Draftwise Launches ‘Legal Ontology’ to Capture The Judgment Layer (Artificial Lawyer)
- Lexis Rolls Out Legal Intelligence Engine Agentic Capabilities (Artificial Lawyer)
