Technology-Assisted Review (TAR), also known as predictive coding or AI-assisted document review, has revolutionized eDiscovery economics. By using machine learning algorithms to identify relevant documents, TAR dramatically reduces the manual review burden that traditionally consumes 60-80% of eDiscovery budgets. Traditional document review requires teams of contract lawyers spending weeks reading through hundreds of thousands of documents at $200-400/hour to identify relevant materials. For a typical case with 500,000 documents, manual review costs $500,000-1,000,000. TAR accomplishes similar accuracy in a fraction of the time and cost. The TAR process works in two main approaches: Continuous Active Learning (CAL) and predictive ranking. CAL systems begin with attorneys reviewing a seed set of documents (typically 200-500), marking them as relevant or not. The algorithm learns from these examples and continuously predicts which documents are likely relevant. Attorneys review the highest-probability predictions, providing feedback that further trains the model. This iterative process continues until the algorithm reaches statistical confidence (typically 95%+ recall). Predictive ranking sorts documents by relevance probability without explicit training, allowing review teams to work from highest-confidence documents downward. Courts increasingly accept TAR as valid discovery methodology, with major litigation adopting these approaches. Key benefits include: 80-90% cost reduction compared to manual review, 50-70% time savings, improved consistency (algorithms don't suffer from reviewer fatigue), and better defensibility (audit trails document the process). Implementation requires careful protocol design—defining the training set, establishing stopping criteria, and documenting the process for potential court challenges. TAR works best for large datasets (50,000+ documents), where the technology shines. For smaller matters, traditional review may be more cost-effective. Leading platforms like Relativity (Assisted Review), Everlaw (AI-Predict), Logikcull (Analytics), and Catalyst incorporate sophisticated TAR engines. Proper TAR implementation requires close collaboration between litigation counsel and technical resources, but the cost savings justify the upfront investment.
Related Articles
