Putting Gen AI to the Test: A Document Review Accuracy Study

Relativity aiR for Review vs. Active Learning

Discover How aiR for Review Can Deliver a 98% Reduction in Review Hours.


Generative AI in legal review has generated both optimism and skepticism – but much of the conversation remains theoretical.

To better understand its real-world impact, Redgrave LLP conducted an independent, head-to-head study comparing Relativity aiR for Review to a managed review workflow incorporating active learning. Designed to be deliberately challenging, the study used a low-richness data set requiring a nuanced legal standard.

The results show that generative AI can meaningfully improve both efficiency and review completeness – while also supporting human decision-making in complex matters.

In this paper, you’ll learn:

  • How generative AI reduced attorney review time by approximately 98% compared to a managed review workflow
  • What measurements like precision, recall, richness, and elusion mean in the context of validating e-discovery workflows, managing risk, and evaluating legal AI
  • How AI-generated reasoning helped experts reassess close-call documents
  • What it takes to test AI against nuanced legal standards
For legal teams evaluating generative AI, the findings offer a detailed, technical look at how AI review tools perform and where they can add the most value.

Download the full study.