Case Study: Kroll reduces first-pass review population by 90% with Relativity aiR for Review
Key results
The challenge
A consumer-sector client faced a dispute involving alleged brand and technology asset misuse, with a large and diverse email data set of about 80,000 documents requiring initial review under tight timelines. An initial attempt at manually coding 22,000 documents proved time-consuming and difficult to scale.
The solution
Kroll used Relativity aiR for Review with a human-expert oversight model, developing and tuning prompts, then validating results against a statistically valid sample designed to achieve a 99% confidence level.
The results, in context
aiR for Review achieved 94.4% recall and delivered an over 90% reduction in the documents requiring full manual review. The human review set was reduced to around 3,500 documents, eliminating tens of thousands of manual review hours.