With speed crucial in document review, AI-powered predictive coding delivers unprecedented speed and cost reductions by using a subset of coded documents to assess the relevance of large pools of uncoded documents.

Challenge

During discovery in a high-stakes case, the customer had to contend with:

• Contentious proceedings in which opposing counsel challenged the initial approach to identifying relevant documents
• The sudden addition of 80,000 documents subject to review
• The prospect of a prolonged manual review of large volumes of similar content
• Inability to accelerate the review due to reliance on a single in-house senior subject matter expert

Solution

• Everlaw predictive coding technology to analyse 80,000 unreviewed documents based on a 5,000-document pre-reviewed seed set
• Flagging of unique content for enriching the model’s training data
• Predictive coding identified 4,471 documents (5.6%) as highly relevant
• Assessment of the predictive model’s precision found 92% precision in detecting 80% of relevant documents
• Accuracy confirmation by subject matter expert to identify discrepancies between reviewer coding and the model’s predictions
• Objective performance metrics, including Precision and Recall, to determine a defensible review completion point