First-pass relevance review burns the most associate hours and produces the most inconsistent results in the entire discovery workflow. Two reviewers with the same protocol routinely tag the same document differently, forcing second-pass review, sampling, and rework. Linear-staffing economics force firms to choose between blowing the budget or cutting corners on accuracy. Relevance Detection attacks both problems at once.
How It Works
The pain today. First-pass review is staffed with contract attorneys at $50–$100 per hour reading documents at 40–60 per hour. The math on a 100,000-document review is brutal: weeks of calendar time, six-figure spend, and a quality ceiling set by reviewer fatigue. Worse, the protocol always evolves mid-review — and re-tagging the documents already reviewed is rarely budgeted.
How DecoverAI solves it. DecoverAI ingests your case definitions — responsiveness criteria, privilege rules, scope parameters — and classifies the entire corpus against the actual legal standards, not just keywords. The AI understands context: a forwarded news article between attorneys is not privileged just because lawyers are on the thread. When the protocol changes, the platform re-runs the classification on every document automatically.
Definition-Driven Classification
The pain today. “Responsive” means something different to every reviewer. One reviewer marks a document responsive because it mentions the contract; another marks the same document non-responsive because the contract is mentioned in passing. The inconsistency forces sampling, QC passes, and disputes — and undermines the firm's ability to certify the production.
How DecoverAI solves it. You provide the definitions for the matter once, and DecoverAI applies them identically to every document in the corpus. Same protocol, same call — every time. Reviewers stop arguing about edge cases and start focusing on the documents that actually need a human judgment call.
Beyond Keywords
The pain today. Boolean search strings either over-collect (drowning the team in noise) or under-collect (missing the smoking gun). Both failure modes are expensive: the first burns associate hours, the second loses the case. The team that built the search string is rarely the team that has to live with its results.
How DecoverAI solves it. DecoverAI uses semantic understanding to identify documents that are responsive even without exact keyword matches, and to exclude documents that contain keywords but are not actually relevant. The system distinguishes between a document discussing legal strategy (privileged) and one that merely mentions an attorney's name in a calendar invite (not privileged).
Scale Without Compromise
The pain today. Linear staffing models do not scale. A 30,000-document review needs weeks of contract attorneys on the clock, an army of reviewers reading at human speed, and a managing associate spending their days fielding tagging questions instead of building the case.
How DecoverAI solves it. Tag 30,000 documents in days, not weeks. In the Tax Credit Investigation matter, DecoverAI processed the entire corpus with consistent tagging in 3 days — work that would have taken traditional review 4 weeks — at a fraction of the cost.