AI-Powered Document Review Workflow Delivered Speed and $13M Savings

The Challenge

A healthcare technology company faced over 7.4M documents following collection, processing and deduplication in a civil litigation. Lighthouse used AI to remove relevance review, reduce privilege review to a fraction of the responsive set, and perform privilege logging and key document identification with minimal linear coding.  

The Solution

The client and counsel opted to use a document review workflow powered by AI to optimize both efficacy and efficiency.  

High-Level Workflow

ROI

Overall, using Lighthouse AI saved $13M.  

AI-Powered TAR

Counsel reviewed less than 5K documents to stabilize, measure, and validate a model that measured 83% precision at 75% recall. This TAR workflow removed 5M documents from review, delivering an ROI of $11M on 1L contract attorney and outside counsel review and QC.  

Non-TAR: 93% Review Reduction

Junk file analysis identified that 1.2M documents out of the 1.3M non-TAR eligible documents were highly likely to be junk and could undergo a sampling workflow instead of a full linear review.  

AI-Powered Privilege Review: 40% Review Reduction

Lighthouse used an AI classifier for privilege for culling and review acceleration. Based on AI results and sampling, counsel determined that 88K documents could be removed without individual review.

Once the privileged documents were identified, Lighthouse used AI to generate first-pass privilege descriptions for the 43K documents in the privilege log. This fully removed 1L human drafting of log lines and instead enabled a quick creation of the log for internal and outside counsel QC.  

Deposition Prep

Using a combination of active learning modeling, categorization analytics, and synthetic documents, Lighthouse surfaced 500 significant documents out of a starting corpus of 1.7M documents that were core to deposition preparation.