LitKit · document review

Document review, end to end.

LitKit is Litco’s document review platform, reinvented around agentic AI. Your familiar e-discovery workflow meets an AI expert that lives in your document repository. Reviewers load their own production or the other side’s, assign batches, code documents, and produce with Bates stamps and a privilege log, with the Matter Agent alongside them the whole way, seeing every document they see. It runs the review, surfaces the documents that decide the case, builds a chronology, and pulls what they need for deposition or trial, learning the case as it goes.

The review grid on the demo matter with tags applied, and the Matter Agent summarizing the finished run in the sidebar
The review grid after a run, with tags applied and the Matter Agent’s summary of the results. Captured from a live instance.

The Matter Agent reviews and tags documents on first pass.

Brief a review the way you would brief your reviewers, with criteria in plain English and tags mapped to your taxonomy. The Matter Agent codes every document, recording the rationale and key quote behind each call. Your team can then spot-check any decision in seconds.

A Review & Tag run in progress. Each square is one document. Squares turn green when tagged and amber when a finding needs a check. Recorded from a live instance.

Criteria in your words.

You write each criterion as a short instruction, the same way you would write one for a person. The run’s tags map to the taxonomy your reviewers already use.

The criteria step of the review wizard, with a criterion named Responsive and its instructions written in plain English
The wizard’s criteria step. Captured from a live instance.

Recall and precision.

When the run finishes, Litco draws a blind sample for reviewers to code and reports recall and precision with confidence intervals—the same statistics courts have accepted for technology-assisted review since 2012. Litco computes those statistics through deterministic sampling, independent of the AI. Reviewers can trace from the audit trail who decided every contested call.

The review run’s results panel: recall and precision with confidence intervals, and the blind sample your reviewers coded
The run’s results panel, with the blind sample your reviewers coded. Captured from a live instance.

Ask about the document you are reading.

Open a document and ask what the thread was about, who else received the attachment, whether anything else in the record contradicts what the witness wrote. The Matter Agent searches the record, pulls the related documents, and cites each one to its Bates number.

The viewer and coding panel, with the Matter Agent citing related documents by Bates number. Captured from a live instance.

Code by voice.

Reviewers can code documents by voice—say “responsive,” “privileged,” or “next document” while reading, and the tags apply. Litco’s own hardware transcribes the audio and never stores the recording.

The foot of the document viewer: the coding panel's tag buttons, with the voice review indicator listening
The coding panel at the foot of the viewer during voice review, with the listening indicator active. Captured from a live instance.

Productions go out with Bates volumes, redactions, and load files. Litco drafts the privilege log, generating an entry for every withheld document with the privilege ground and a description drawn from the document and from its metadata. Counsel can review the full log before service.