Generative AI for Efficient Document Review in High-Stakes Litigation

A Fortune 500 food company found itself in a complex legal battle involving a vendor’s software. As new counsel came on board, they faced a massive document review challenge: approximately 230,000 legacy documents needed evaluation, and the team also had to address 55 new Requests for Production (RFPs).

Traditionally, this could demand 20 to 30 attorneys working around the clock for two weeks. Instead, after consulting with counsel, IntrepidX proposed a more efficient approach powered by a combination of the latest generative AI technology and traditional Technology-Assisted Review

Our Solution

Phase 1: Generative AI-Powered RFP Tagging
To tackle 55 RFPs, we consolidated them into 35 subject areas and created a “spray prompt,” guiding the AI to categorize documents by relevance. Using Relativity aiR, the entire corpus was processed in a single pass—23 hours of machine time—to identify 6,000 highly responsive documents and 12,000 borderline ones. To refine the borderline group, targeted keyword searches (e.g., “crash,” “malfunction,” “failure”) helped defensibly remove items over-tagged due to ambiguous phrasing (e.g., “issues with software”).

Phase 2: Adaptive TAR Modeling
Building on insights from Phase 1, we blended 500 highly responsive documents with a random sample of 500 new documents to form a seed set. Attorneys validated this set, then used it to train a Technology-Assisted Review (TAR) model. Because the model incorporated prior AI findings, there was no need for additional prompts or extra Relativity costs. The TAR system quickly prioritized new document collections, eliminating the need for large review teams.

Impact

  1. Cost Savings: Reduced reliance on extensive review teams.
  2. Time Efficiency: Tasks that once required weeks were completed in hours.
  3. Defensible Process: Rigorous validation and iterative refinement met legal standards.
  4. Scalability: The AI-driven workflow easily adapted to new document sets.


Conclusion

By combining human oversight with generative AI and TAR, IntrepidX delivered a faster, more cost-effective, and defensible review process. This approach not only met urgent production deadlines but also set a precedent for handling large-scale litigation. As the industry faces increasing data volumes, this blend of advanced AI and expert validation is poised to become the new standard for efficient, high-stakes document review.