Investigations at the speed of a prompt

Internal investigations have long been defined by the wait between question and answer. In our ongoing series on the Future of eDiscovery, we have tracked the move upstream toward early case assessment and the embedding of trusted AI across the discovery lifecycle. That shift is especially powerful in investigations, where generative AI is closing the gap, turning reactive scrambles into controlled, iterative, and increasingly self-service investigations.

Current state of electronic investigations

Traditional investigations involving email, communications, and documents run on legwork. A typical matter involves multiple teams that collect the data, cull the data, negotiate keywords, and ship documents to an outside review team—then wait sometimes days or weeks for a memo. Every new question can restart the loop. For allegations that need triage in hours, or cross-border matters spanning multiple languages, that linear model cannot keep pace with modern data volumes or urgent demands.

The loop, replaced by a conversation

Our Relativity aiR Investigations approach, built on RelativityOne and its aiR Assist capability, replaces that loop with a conversation. An investigator asks the corpus a question in plain language and receives a cited, source-linked answer in minutes—then drills in, refines, and asks the next question in the same session. Legal, compliance, and HR teams can drive it directly, while KPMG LLP manages the environment, handles collection and processing, and provides prompt engineering, quality control, and validation.

Answers arrive in minutes, not weeks; investigations run on demand, without an outsourced review cycle; and every answer is cited back to source documents—so conclusions can be traced, tested, and produced.

Real-world impact: aiR assist in action

The impact is easiest to see in the numbers. In one recent engagement, aiR Assist helped filter roughly 724,000 documents down to 31 critical findings over a single weekend.¹ In a separate multilingual investigation, approximately 50,000 documents across multiple languages were assessed in days rather than weeks.² Both matters drew on the same workspace, where three aiR tools operate in a single workflow: Assist for conversational, cited interrogation of a document set; Review for responsiveness and privilege calls; and Case Strategy for narrative development.