Since the release of ChatGPT, generative AI has become a buzzword in the legal tech community. Many of us began asking key questions: How will it impact our workflows? How accurate is it? Is it safe and secure? These questions sparked curiosity, which in turn has driven innovation. If one thing became clear after Relativity Fest 2024, it’s that Relativity has been quick to capitalize on generative AI’s potential with Relativity One.
Their first AI-driven product, aiR for Review, launched just weeks ago and was a central focus at Relativity Fest 2024. Two other AI tools—aiR for Case Strategy and aiR for Privilege—are slated for release later this year and early next year. Based on early results, the legal community’s adoption of generative AI may happen sooner than anticipated. In some use cases, aiR for Review is showing a Precision rate of over 90%, Recall above 70%, and cost savings of 50% or more. In one instance, a single attorney reviewed 1,000,000 documents in just 18 hours. However, to truly harness the power of aiR, teams will need skilled Consultants and Project Managers to implement, monitor, and fine-tune the tool effectively.
aiR for Review is a valuable addition to any case team. Cases previously suitable for TAR might now be ideal candidates for aiR. It can also be used alongside Active Learning, enhancing efficiency. While Active Learning isn’t going anywhere, aiR will reduce first-level review time and costs, shifting more focus to prompt engineering—the process of drafting review protocols, addressing document context, and managing coding calls.
Results depend on three key factors: developing prompts, refining the review process, and validating outcomes. As case teams gain familiarity with aiR, accuracy will improve. Success will require training, trial and error, and leveraging legal service providers with expertise in managing aiR projects.
Yes, Relativity’s aiR for Review uses Microsoft Azure’s Open APIs. Data is not retained by Microsoft or any third party, and there’s no connection to open-source internet. Documents are sent to the model for analysis, returned to Relativity, and then immediately deleted from Microsoft Azure.