Undisclosed

During a high-stakes discovery matter, the client faced contentious proceedings where opposing counsel challenged the document identification approach. The sudden addition of 80,000 documents requiring review, combined with reliance on a single in-house senior subject matter expert, made accelerating a prolonged manual review of large volumes of similar content effectively impossible.

Elevate deployed Everlaw predictive coding technology to analyse 80,000 unreviewed documents using a 5,000-document pre-reviewed seed set. The model flagged unique content for training enrichment and identified 4,471 documents as highly relevant. A subject matter expert confirmed accuracy, and objective Precision and Recall metrics were used to determine a defensible review completion point, achieving 92% precision in detecting 80% of relevant documents.

The predictive coding approach reduced the document review population by 90% and cut estimated review hours from 1,500 to 100 — a 93% reduction — dramatically accelerating the discovery timeline. Integrated quality control processes also reduced errors stemming from inconsistent human coding practices.

• A small seed set (5,000 documents) is sufficient to train a model that reliably triages 80,000+ documents, enabling defensible TAR workflows.
• Objective performance metrics (Precision and Recall) provide the evidentiary basis needed to withstand opposing counsel challenges.
• SME involvement for accuracy confirmation remains essential even when AI handles bulk classification.