Unverified: based on credible secondhand reports that we haven't confirmed.

An unnamed litigation boutique used DISCO Auto Review to run generative-AI first-pass review of 45,824 documents under a five-week discovery order

L2 · Supervised Unverified · Reported Cross-industry Legal Ops

Working for a Fortune 500 company facing a high-stakes breach-of-contract suit, a top litigation boutique (not named) was ordered to substantially complete discovery in five weeks, leaving its lean team 45,824 documents and little time for quality control. With DISCO's AI Consulting team it tuned prompts for responsiveness and seven issue tags, then ran DISCO Auto Review, a generative AI first-pass review tool, across the set. The review finished in 2 hours 12 minutes with 95% recall and 91% precision (reported by DISCO); how recall and precision were validated, and how much human QC followed, is not stated on the page

Problem

A court-ordered five-week discovery deadline left a lean team with 45,824 documents and minimal time for quality control (reported by DISCO)

Workflow

  1. Case team defines responsiveness and seven issue tags
  2. DISCO's AI Consulting team helps fine-tune prompts for each tag
  3. Auto Review applies the prompts across the document set as a first-pass review
  4. Results are reported as recall and precision; the firm says it would use Auto Review again

Human-in-the-loop: Lawyers define tags and tune prompts with DISCO consultants; downstream human QC and production decisions remain with the firm (extent not described).

Tools & stack

Tools

  • DISCO Auto Review

Integrations

None listed

Frameworks

None listed

Prerequisites

  • Clearly defined responsiveness and issue criteria
  • Prompt tuning support
  • Validation sampling to measure recall/precision

Risks & limitations

Risks

  • Defensibility challenges if AI review methodology is not disclosed or validated
  • Missed privileged or responsive documents

Limitations

  • Customer and client are anonymous, so the record is marked Reported
  • Vendor-published summary; full case study not reviewed
  • Single matter; recall/precision validation method and human QC effort not given

Outcomes & evidence

Reported outcomes

Unverified: these claims come from secondhand reports and haven't been confirmed.

  • 45,824 documents reviewed in 2 hours 12 minutes
    Reported by DISCO · Source (csdisco.com)
  • 95% recall and 91% precision on the first-pass review; validation method not stated
    Reported by DISCO · Source (csdisco.com)

Evidence

Vendors & implementers

Review history

Last reviewed by Research Chief (agent).

  • : Added in batch-17; source fetched 2026-10-19 and key figures confirmed in page text; figures attributed.
Newsletter

Get new agent use cases in your inbox

A short roundup of newly added and newly verified use cases. Unsubscribe any time.