Lighthouse Delivers $20M Savings in Fast Paced Second Request

Challenge

Antitrust regulators issued a broad, high-stakes HSR Second Request to investigate a global company’s high-profile acquisition. To comply, the company and their outside counsel were faced with analyzing 30+ TB of data in less than a month.

Solution  

Lighthouse developed an AI driven approach powered by Lighthouse’s proprietary Large Language Models (LLMs) to eliminate relevance review, substantially reduce privilege review, perform privilege logging and assemble the names list, and identify key documents to mitigate risk without linear review.

Simultaneously, Lighthouse operational teams executed multiple custom workflows to process and produce over 10TB+ of data and 20M+ images in 3 weeks—flawlessly.

Their work included building a custom linking workflow for M365 cloud attachments and a secure data repository for work product reuse in a related antitrust litigation, while ramping a 300-person linear review team via Lighthouse’s Managed Review solution to meet the aggressive production deadline.

Lighthouse AI Savings & ROI Breakdown

Overall, using Lighthouse AI saved $20M+ with the following workflows:

• TAR powered by Lighthouses AI proprietary LLMs
• Privilege identification powered by Lighthouse AI proprietary LLMs
• Privilege log and names list generation via Lighthouse AI proprietary LLMs
• Key document identification powered by linguistic modeling and AI
• Junk file analysis
• Cross matter analytics

Highlights of each of these workflows and the associated ROI are below.  

Lighthouse’s Proprietary Predictive AI for Relevance: 40% Narrower Responsive Set Than Other TAR Tools

Counsel only needed to review around 4,000 documents to stabilize and validate a TAR model backed by Lighthouse’s predictive LLMs that measured 85.91% precision at 76.49% recall. Lighthouse’s predictive AI model for relevance is shown to deliver a 30-40% smaller and more accurate R-set based upon comparative bake-offs. Had Lighthouse implemented a Second Request workflow using commercially available TAR tools, we would have likely seen a much broader responsive set, thereby requiring additional privilege review and production.

Specifically, if Lighthouse had used traditional TAR tools, it would have resulted in an estimated additional 2M documents in scope for production, translating into approximately $5.8M in additional privilege review costs and $400,000 in added production expenses—costs avoided by using Lighthouse Responsive AI.

Separately, Lighthouse’s Review Management team performed Junk File analysis and identified that around 95,000 documents out of the more than 860,000 non-TAR eligible documents were highly likely to be junk and could undergo a sampling workflow instead of a full linear review.

Privilege Review: 45%+ Reduction  

Lighthouse built an AI model for privilege that both removed documents from privilege review and accelerated the remainder. Using a tiered approach, counsel determined that around 780,000 documents (out of the more than 1.7M eligible for privilege review) could be removed from privilege review without linear review because they fell below the cutoff score and were unlikely to be privileged.

Once the privileged documents were identified, Lighthouse’s generative AI was used to create first-pass privilege descriptions for the more than 100,000 documents on the privilege log. This removed the need for human drafting of log lines. As a result, the log required an investment of mere hours as opposed to days and the heavy expensive of a full contract attorney review and outside counsel QC.

Lighthouse also used generative AI to build the names list for the privilege log. Lighthouse’s AI model quickly analyzed around 130,000 documents to identify and provide close to 15,000 normalized names with titles for the privilege log.

Key Document Identification

The Lighthouse AI team used modeling to quickly surface documents that could pose an antitrust risk to the company. This process eliminated the need for the more dated approach of search+linear review for key documents and issues tags. Out of an initial tranche of 4.4M documents, the Lighthouse team identified roughly 270 documents of the greatest interest and sensitivity and delivered them to counsel in 5 deliveries over the course of just 3 weeks.

To provide additional support to counsel, the Lighthouse AI team also categorized the final Responsive document set based on risk profile—classifying a broad set of documents as Likely Risky and Likely Safe over the course of just 1.5 weeks.

Work Product Reuse for Cross-Matter Efficiency

The Second Request had a large set of overlapping data with a concurrent antitrust litigation. To ensure there was no duplicative review and to drive consistency, Lighthouse built a secure data repository that enabled work product reuse between this Second Request and that concurrent litigation and used Lighthouse AI to drive cross matter analytics. With Lighthouse AI, Lighthouse repurposed calls for around 680,000 documents, resulting in a savings of $3.5M between first pass review, 1L QC, and outside counsel QC.

Speed, Quality, and Operational Delivery  

The scale and complexity of this Second Request required an extraordinary cross-functional effort across several Lighthouse Client Services and Operational teams who worked tirelessly to deliver a seamless, high-quality result under immense time constraints. The core of this matter was completed in under 60 days, demonstrating an exceptional level of execution.

Key highlights of the operational delivery include:

• High-Volume Processing and Custom Workflows: Lighthouse processed over 30TBs of data, ensuring rapid ingestion, indexing, and AI-powered classification. A custom workflow was implemented to link M365 cloud attachments, meeting regulatory requirements.
• Flawless Data Production at Scale: Within 3 weeks, the Lighthouse team produced over 10TBs of data and more than 20M images, achieving a 100% error-free production result. This ensured compliance without delay or rework.
• Scalable Review Team for Complex Work Streams: Leveraging our Managed Review capabilities, Lighthouse rapidly scaled a linear review team to 300 professionals. The team expertly navigated multiple complex work streams, dynamically segmenting data to mitigate risk while accelerating the review process to meet the production deadline. The bulk of review ramped and concluded in a mere 4 weeks.

This work showcases the coordination, detail, operational excellence and sheer dedication of the Lighthouse team in delivering a timely, high-quality outcome in an intense regulatory investigation.