Why Build Your Own AI Model When Frontier Models Already Exist?

Thomson Reuters has developed a domain-specific frontier AI model, "Thomson," to enhance its legal, tax, and compliance software offerings. By utilizing open-weight checkpoints like Qwen 3.5, the company avoids the limitations of basic fine-tuning APIs, instead employing a rigorous pipeline of continuous pre-training and reinforcement learning. This strategy leverages the company’s vast proprietary data and the expertise of over 3,000 internal professionals to ensure accuracy and mitigate catastrophic forgetting. The model functions within agentic workflows, such as the CoCounsel platform, to automate complex tasks like M&A due diligence and legal research. With a total investment of $40 million, this approach prioritizes sovereign control over the AI roadmap, allowing the company to integrate domain-specific judgment into its software while maintaining the flexibility to adopt future open-weight advancements.