E-Discovery Triage: Using AI to Prioritize Document Review Queues

Document review in complex civil litigation can cost hundreds of thousands of dollars — sometimes millions — in attorney time reviewing large collections of electronically stored information (ESI). Technology-Assisted Review (TAR), also called predictive coding, uses AI to rank documents by likely relevance, allowing human reviewers to focus on high-priority items. Courts in the U.S. and UK have accepted TAR as a defensible review methodology when properly implemented.

How TAR works

1. 1A seed set of documents is reviewed and coded by a senior attorney (relevant/not relevant)
2. 2The TAR system learns from the seed set and ranks the remaining documents by predicted relevance
3. 3Reviewers focus on the high-ranked documents first, working down the ranked list
4. 4Validation samples confirm the model's accuracy at stopping points
5. 5Documents below a defensible relevance threshold may be set aside without full human review

The big idea: AI relevance ranking cuts the human review burden dramatically — but the methodology must be defensible and attorney-supervised.