What is AI eDiscovery (TAR, Predictive Coding)?
TL;DR
Legal AI for document review, privilege detection, TAR 2.0 continuous active learning and generative AI review — delivering review time -70%, outside counsel cost -50% and 98% privilege accuracy. Relativity aiR/Everlaw/Reveal/DISCO power a $25B market by 2030.
AI eDiscovery (TAR, Predictive Coding): Definition & Explanation
AI eDiscovery (Electronic Discovery, Legal Hold, Document Review, TAR Technology Assisted Review, Predictive Coding, Generative AI Legal Review) integrates Identification (custodian/source identification across Microsoft 365/Google Workspace/Box/Slack/Teams), Preservation (legal hold notice issuance, acknowledgment tracking, auto-release), Collection (forensic image, chain of custody, hash verification, mobile/chat/cloud), Processing (de-duplication, email threading, modern attachments, OCR, metadata extraction), Review (TAR 2.0 continuous active learning, generative AI review, privilege detection, issue coding), Production (Bates numbering, PII/PHI redaction, privilege log, load file) and Presentation (trial prep, StoryBuilder, exhibit management). It unifies the full EDRM workflow, is FRCP/Sedona Conference compliant, and supports 40+ languages (CN/JA/DE/ES/PT/RU/AR). Market $15B (2024) → $25B (2030) at 10% CAGR. Magic Quadrant leaders: Relativity, Everlaw, Reveal, DISCO, KLDiscovery. Leading platforms: (1) Relativity aiR ($3.6B, 300,000+ users, 98% of Am Law 200, aiR for Review/Privilege/Case Strategy, generative AI pioneer, $50K-5M/yr); (2) Everlaw ($2B, 2,000+ firms, Skadden/Latham/Quinn Emanuel, cloud native, StoryBuilder trial prep, $30K-2M/yr); (3) Reveal Brainspace ($1.5B, 200+ large firms, concept search/cluster wheel, $50K-3M/yr); (4) DISCO (NYSE:LAW $1B, Cecilia AI, 1,500+ customers, $30K-1M/yr); (5) Logikcull by Reveal (SMB self-service, $25-300/GB/mo); (6) Lighthouse Discovery ($500M, managed service + tech, Fortune 500, $200K-5M/yr); (7) KLDiscovery Nebula ($1B, global 20+ countries, $100K-3M/yr); (8) Casepoint ($200M, government/CCPA/GDPR, $80K-1M/yr); (9) Onna ($50M, Slack/Teams/modern source, $30K-500K/yr); (10) Hanzo ($30M, web/social collection), Exterro/Consilio/CS Disco/Opus 2/Nuix and Japan's UBIC Lit i View / FRONTEO KIBIT. Key use cases: TAR 2.0 continuous active learning (Relativity Active Learning, 80%+ recall / 70%+ precision targets, Sedona Principle 11 compliant, court-defensible); generative AI legal review (Relativity aiR for Review, Everlaw AI Assistant, DISCO Cecilia for document summary + issue coding + privilege detection; review time -70%); privilege detection (aiR for Privilege, predictive privilege, 98% accuracy, auto privilege log, attorney work product protection); legal hold automation (custodian interview automation, hold notice issuance, acknowledgment tracking, release, audit trail, Zubulake standard); chain of custody (Collection→Processing→Review→Production with hash, forensic images, FRE 901 admissibility); modern data collection (Slack/Teams/WhatsApp/iMessage/mobile forensic via Cellebrite, cloud Microsoft 365/Google Workspace); concept search/cluster (Brainspace concept wheel, topic modeling, investigation triage 3 days → 4 hours); StoryBuilder trial prep (Everlaw witness/exhibit/issue timeline, trial notebook, deposition outlines); MDL review (multi-firm coordination, tag sharing, unified production); government/SEC/DOJ investigations (Casepoint FedRAMP/CJIS, large-scale data production, subpoena response). Results: Relativity 300,000+ users, Everlaw 2,000+ firms, Reveal 200+ large firms; document review time -70% (1,000→300 hrs/matter), outside counsel cost -50% ($500K→$250K), privilege accuracy +30% (85%→98%), custodian interview time -60%, trial prep time -40%, investigation triage 3 days → 4 hours, 80%+ recall/precision; market $15B (2024) → $25B (2030); ROI 5-20x. Key considerations: TAR defensibility (Sedona Principle 11, statistical sampling, court-defensible, Da Silva Moore precedent, cooperation doctrine); generative AI hallucination risk (citations mandatory, document links, human review, sampling QC, avoid raw OpenAI/Anthropic API for production review); privacy/PII redaction (GDPR/CCPA/HIPAA, automated redaction verification, re-identification risk); cost management (per-GB hosting, per-document review, linear vs TAR cost comparison, early cost estimate, phased production); court/protective order compliance (production format, confidentiality designation, FRE 502(d) clawback). 2026 trends: generative AI legal review goes mainstream (Relativity aiR for Review, Everlaw AI Assistant, DISCO Cecilia accelerate review 50%+ using GPT-4/Claude via Bedrock); cloud-native shift (RelativityOne, Everlaw cloud-native, Reveal cloud cut implementation from 3 months to 2 weeks); modern data sources (Slack/Teams/WhatsApp/iMessage/Salesforce/Notion via Onna chat threading); privilege AI (aiR for Privilege at 98% accuracy, auto privilege log, work product protection); cross-border eDiscovery (GDPR Schrems II, data residency EU/UK/APAC via Casepoint/Relativity Government); investigation/compliance integration (SOX/FCPA/UK Bribery Act, Reveal Brainspace + Onna, 3-day internal investigation triage); Japan readiness (Relativity One + UBIC Lit i View + FRONTEO KIBIT for US discovery and cross-border investigation).