Litigation collections are messy: mail exports, office documents, scanned paper, phone photos, voice memos, and archives inside archives. The platform ingests all of it into a single unified search index, so no class of ESI stays dark - and nothing gets quietly left behind.
Email Archives, Done Right
Email is usually the center of gravity of a collection, and it is where processing shortcuts hurt the most. The platform ingests full mailbox exports (mbox) and individual .eml files with a streaming parser that handles multi-gigabyte mailboxes without breaking a sweat.
- Conversation threads are reconstructed, and thread identity survives subject-line rewrites.
- Folder labels are preserved exactly as the custodian filed them.
- Duplicate copies of the same message - inbox, sent, labels - collapse into one record while every copy's provenance is retained.
- Every attachment is extracted, hashed, deduplicated, and indexed as its own document, linked back to every message that carried it.
Documents, Spreadsheets, and Presentations
PDF, Word, Excel, PowerPoint, CSV, HTML, plain text, and Markdown are parsed natively. Each file's own dates are preserved and become searchable, filterable metadata - because in discovery, when a document was made matters as much as what it says.
Scanned Paper, Images, and Recordings
The classes of ESI that most platforms leave unsearchable are handled as first-class citizens:
- Scanned and legacy documents. Files with no text layer are recovered by AI-powered OCR. Long scans of 100+ pages are split and processed automatically, so even a fully scanned court filing ends up word-searchable.
- Images. Every image gets an AI vision pass that produces a written description plus a verbatim transcript of any visible text - screenshots, photographed documents, whiteboards, charts. Phone formats such as HEIC are converted automatically.
- Audio and video. Recordings are transcribed automatically with speaker separation and timestamps, with automatic language identification. A conversation in a voice memo becomes a searchable, quotable transcript.
Archives and the Long Tail
ZIP files - including archives nested inside archives - are unpacked, and every internal file is catalogued individually with its own hash and its original modified date, then routed to the right processing pipeline. Anything the platform cannot parse is inventoried and explicitly marked unsupported rather than silently dropped, so the index is complete and provably so.
One Index for Everything
Every processed item - an email body, a spreadsheet, the transcript of a voicemail, the text inside a screenshot - lands in the same searchable space. A single query sweeps the entire collection at once, regardless of what form the evidence originally took.
Works Together With
- Hybrid SearchSemantic and keyword search fused and reranked, with the custodian, date, and issue filters that discovery questions actually take.
- Chain of CustodyImmutable originals, content hashing, full lineage, inventory sign-off, and append-only audit trails throughout.
- Intake & OperationsSecure browser uploads, automatic ingestion of new arrivals, a missing-records register, and live pipeline monitoring.
Frequently Asked Questions
Can you make scanned paper and photographed documents searchable?
Yes. Scanned PDFs and legacy formats with no text layer are recovered with AI-powered OCR, and every image gets an AI vision pass that produces a description plus a verbatim transcript of any visible text. Long scans of 100+ pages are split and processed automatically.
What happens to files the platform cannot parse?
They are never silently dropped. Every unsupported file is inventoried and explicitly marked, so the review team always knows exactly what is and is not in the index - and can make a defensible record of it.
Do duplicate documents get reviewed twice?
No. Every item carries a content hash, so identical files and duplicate email copies collapse into one reviewable record - while the provenance of every copy is preserved. The same attachment sent five times is reviewed once, with all five parent links intact.
Can the platform handle collections in multiple languages?
Yes. The corpus can mix languages freely. Semantic search is language-agnostic, and queries are translated automatically for keyword matching, so a search in English finds relevant documents written in another language.