Robert Keeling & Ray Mangum: Results Are In: 5 Lessons From an Independent Study of aiR for Review

Key Takeaways

• Relativity aiR matched traditional review accuracy on complex compliance set
• AI workflow cut first‑pass review time by roughly 40%
• Human reviewers still needed for final judgment on nuanced documents
• Cost per document reduced due to faster AI processing
• Study underscores importance of training data quality for generative AI

Pulse Analysis

Generative AI has moved from speculative buzz to practical application in e‑discovery, but many firms remain skeptical about its ability to interpret nuanced legal standards. Redgrave LLP’s independent study addressed that gap by pitting Relativity’s aiR for Review against a seasoned active‑learning workflow on a 45,000‑document set drawn from public data. The chosen documents required reviewers to assess compliance with federal regulations on pharmaceutical marketing and controlled substances, a task that demands contextual judgment far beyond simple keyword detection. By designing a test that mirrors real‑world litigation demands, the study provides a credible benchmark for AI’s true capabilities in legal review.

The results were striking. While both methods achieved similar precision in identifying responsive documents, the AI‑driven workflow completed the first‑pass review roughly 40% faster, translating into significant labor savings. Cost per document fell as the AI processed large volumes with minimal human intervention, yet the study confirmed that final validation by experienced attorneys remained essential for nuanced determinations. This hybrid model—AI for rapid triage, humans for final judgment—delivers a balanced approach that mitigates risk while capitalizing on efficiency gains.

Industry implications are profound. Law firms can now justify investing in generative AI platforms like Relativity aiR, knowing that performance does not degrade on complex compliance issues. However, the study also highlights the critical role of high‑quality training data; AI accuracy hinges on the relevance and representativeness of the documents fed into the system. As the technology matures, firms that integrate AI thoughtfully—pairing it with expert oversight and robust data governance—are poised to gain a competitive edge in cost‑effective, high‑quality e‑discovery services.