Questions worth a straight answer.

It takes a set of legal documents, turns them into structured and searchable text, then runs agents over them to produce a first-pass review. Depending on the workflow, that means clause extraction against your playbook, risk flags on deviations, diligence triage, or a privilege screen. Output arrives as ranked findings, each carrying a citation to the page and clause it came from.

Most embedded assistants answer questions about a document you have already opened. DocuLightning works the other way around, running an unattended pass across the whole set and telling you which documents deserve opening. It also plugs into your DMS rather than replacing it, so nothing moves out of iManage or NetDocuments permanently.

PDF including scanned and image-only, Word, Excel, PowerPoint, email including PST and MSG with attachments, TIFF, images, plain text, and RTF. Handwriting, stamps, redlines, faxes, and multi-column layouts are handled by the vision extraction models rather than an old-style OCR engine, which is what makes legacy archives workable.

On contract sets with an established playbook, our design partners have seen 94% of first-pass review complete without a human touch. Accuracy drops on degraded scans, unusual jurisdictions, and bespoke drafting, which is exactly where the system escalates instead of guessing. We publish confidence scores rather than a single marketing number because the number depends entirely on your documents.

The honest framing: this replaces the first read, not the judgment. Every finding is presented for sign-off, and the audit log records who signed.

Less than an unconstrained chat assistant, and not never. Findings must be grounded in a retrieved passage with coordinates, and a verification pass re-checks each citation against the source page before delivery. Anything that fails verification is escalated rather than shipped. Courts have been consistent that the filing lawyer bears responsibility regardless of the tool, and the product is designed around that reality rather than against it.

Self-serve trial is about ten minutes: sign up, drag in documents, run. A DMS integration takes a few hours with your IT team. Loading a playbook takes an afternoon if your fallback positions are already written down, and a week if this is the exercise that finally makes you write them down.

Your standard positions and acceptable fallbacks, clause by clause. Cap liability at 12 months of fees, no automatic renewal beyond 12 months, no assignment without consent, and so on. Upload an existing one as a document, or build it in the interface. The agents mark every incoming paper against it and tell you what deviates and by how much.

iManage, NetDocuments, SharePoint, Relativity, Box, and Google Drive on Practice and above, plus a REST API and webhooks for anything else. Enterprise customers get help wiring up whatever internal system your practice actually runs on.

Usually legal ops or a litigation support lead runs the ingestion, associates and contract managers work the findings queue, and a partner or GC signs off. Unlimited seats on every plan exist so the reviewer checking the work is never a line item.

Yes, on Enterprise. Single tenant or a deployment inside your own VPC where document content never leaves your AWS account. Model inference in that configuration runs against self-hosted open-weight models.

Clients are already declining to pay associate rates for a first read that a machine can do overnight. Our design partners have generally moved that work to fixed fee and put the recovered hours into the parts clients still pay a premium for. We would rather say that plainly than pretend the pricing model is unaffected.

A small team out of Chicago with backgrounds in enterprise infrastructure, document machine learning, and litigation support. We take design partners before we take large contracts, and we would rather lose a deal than oversell what the system does.