Thomson Reuters today began testing a custom version of OpenAI's latest language model in its legal assistant CoCounsel. The implementation marks the primary corporate adaptation of the o1 mini model and shows how major corporations are actually transforming their artificial intelligence strategies.
The media and technology giant has implemented a strategic approach by deploying specialized AI models from OpenAI, Google and Anthropic, each optimized for specific legal tasks. Industry analysts imagine that this strategy, combined with o1-mini's novel features, could turn out to be a blueprint for the usage of AI in corporations across industries.
“Each model – OpenAI, Google Gemini and Anthropic – brings unique capabilities tailored to the needs of specific workflows,” said Joel Hron, chief technology officer at Thomson Reuters, in an exclusive interview with VentureBeat.
Based on these competencies, the corporate carries out various legal tasks. “OpenAI focuses on generative tasks reminiscent of summarization and conversational AI inside CoCounsel. Google's Gemini is good for long-context tasks and enables deep integration into large legal documents. Claude from Anthropic is concentrated on workflows that require high sensitivity and customization, reminiscent of: B. Tax and compliance use cases.”
According to James Dyett, Head of Platform Sales at OpenAI, who also spoke to VentureBeat in an exclusive interview, the brand new o1-mini model significantly improves AI follow-up capabilities.
“OpenAI o1-mini is designed for workflows that require professionals to discover very minor but potentially consequential terms and errors in legal briefs,” said Dyett. “Compared to GPT-4, OpenAI o1-mini has been trained to spend more time considering through legal complexities.”
AI is showing great progress in analyzing legal documents
Initial testing has shown significant performance improvements in real-world applications. Hron pointed to specific examples from her evaluation process.
“In our testing of o1-mini to detect privileged emails, the model demonstrated a remarkable ability to discover situationally nuanced instances of privilege that had previously eluded even powerful models like GPT-4,” he said. “This advancement is a direct reflection of o1-mini’s improved reasoning and contextual understanding.”
The strategy has produced significant results. Thomson Reuters reports a 1,400% increase in CoCounsel users over the past yr. The system has transformed several key legal workflows, particularly in document management and evaluation.
“Document review, legal research, drafting and revision have all experienced significant improvements,” Hron noted. “These improvements have increased productivity and permit lawyers to concentrate on higher-value tasks.”
From AI customer to AI developer: The strategic expansion of Thomson Reuters
The company's AI strategy goes beyond leveraging existing technology. Thomson Reuters recently acquired the UK-based company Secure drawing technologiesa specialist in legal language models, marks a big step in AI development.
“Our technique to develop proprietary LLMs through Safe Sign Technologies complements our partnerships by giving us greater control over data security, customization and value effectiveness,” explained Hron. “It allows us to more directly leverage our biggest strengths – our proprietary content and world-class domain experts – to create unique solutions that only we are able to deliver.”
Managing multiple AI models required sophisticated infrastructure support. Thomson Reuters partnered with Amazon Web Services to handle the computational burden and have become considered one of the primary customers of AWS Sagemaker HyperPod.
“We have close and long-standing relationships with all of those providers and have the computing infrastructure obligatory to satisfy the demand for every of those models,” Hron said. “This allows us to truly optimize costs by strategically assigning tasks to the suitable model.”
The development has caught the eye of each technology leaders and investors. Dyett highlighted the broader implications for the usage of AI in businesses.
“OpenAI is working with large corporations to know the ways through which frontier models reminiscent of o1-mini or customized versions of o1-mini can support specific use cases,” he said. “These findings allow us to enhance our modeling capabilities and discover additional legal tasks suitable for adapting the OpenAI o1-mini reasoning.”
While enterprise AI has traditionally focused on broad functionality, Thomson Reuters' implementation of o1-mini signals a critical shift toward precision-engineered models that excel at highly specialized tasks.
The model's ability to capture nuanced legal distinctions that even GPT-4 missed suggests that the longer term of AI lies not in jack-of-all-trades systems, but in sophisticated networks of specialised models working together.
For the legal industry, where a single missed detail can have thousands and thousands of dollars in consequences, this precision approach could redefine the standards for AI use.
