Skip to content

AI Investigate

When you ask your AI assistant to find or analyse documents through Aurora, you’re not running a keyword search. You’re invoking AI Investigate — Consilio’s enterprise platform that combines retrieval-augmented generation with agentic search to investigate matter data on your behalf.

AI Investigate is built for the way real legal investigations actually work — open-ended questions, evolving leads, evidence that has to hold up to scrutiny.

  • Retrieval-augmented generation. Every answer is grounded in your actual documents. The model only writes what the evidence supports, and every claim is cited back to its source.
  • Agentic search. AI Investigate doesn’t stop at a single query. It iterates — refining searches, following leads, and pulling in additional context until it has enough material to answer the question well.
  • Enterprise-grade. Designed for the volumes, sensitivity, and compliance demands of large-matter eDiscovery work.

When the answer matters more than the keyword match, this is the engine that produces it.

Aurora gives MCP-aware AI assistants two tools that map onto AI Investigate:

ToolWhat it does
searchWorkspacesLists every workspace inside a matter, so the assistant can pick the right scope.
searchDocumentsAsks AI Investigate to investigate a question inside a chosen workspace.

When a matter has exactly one workspace, searchDocuments auto-selects it and runs the investigation immediately. When a matter has several, the tool returns the workspace list first so the assistant can confirm which one you meant before AI Investigate spends cycles on it. This keeps results tightly bounded and prevents accidental cross-workspace investigations.

  • “What does the workspace say about the SEC subpoena timeline?”
  • “Summarise the key documents about the settlement letter in matter A12345.”
  • “Find every mention of ‘material weakness’ across this workspace and tell me where they appear.”

AI Investigate returns answers in natural language with embedded citations — document title, identifier, and a snippet of the matching passage — so your assistant can hand them straight back to you with clickable references. Nothing is summarised in the abstract; every finding traces back to the underlying file.