For legal & eDiscovery teams

Your evidence lives
in four tools. We built one.

Legal hold, review, and production run across disconnected systems while courts rewrite the rules for AI in discovery. OrcheSight keeps them in one case record, with AI on your own infrastructure.

The current state

Legal hold, review, and production still live in separate systems

A fragmented toolchain

Legal hold, collection, review, and production run in separate systems. Each handoff means re-uploading data, reformatting exports, and a harder chain of custody to defend.

Privilege review by hand

Associates and contract reviewers page through documents one at a time to catch privilege and confidential material, knowing one miss becomes a clawback motion.

Uncertainty about AI tools

Bar guidance and protective order language are still changing. Teams lack a clear answer for what AI use is permitted on privileged material, or how to document it.

Why this matters now

The rules for AI in discovery are being written now

Two recent decisions reshaped how legal teams can use AI in discovery: a federal court accepted a generative AI review workflow, and a separate court tightened what AI tools are permitted on discovery material under a protective order.

Schulte v. LinkedIn Corp.

U.S. District Court, N.D. Cal. (Magistrate Judge Laurel Beeler) · June 30, 2026

The court accepted a generative AI workflow to make final responsiveness determinations in discovery, treating it like established technology-assisted review and declining to require extensive disclosure of the underlying validation metrics.

The first federal decision accepting generative AI for final responsiveness determinations, under continued attorney supervision, not in place of it.

Jeffries v. Harcros Chemicals

U.S. District Court, District of Kansas

The protective order was amended to bar uploading any discovery material, confidential or not, to open or public AI tools. Closed, secure AI remains permitted for confidential material only under contract: no training on inputs, no onward disclosure, deletion on request.

Protective orders are starting to name AI architecture directly — where a model runs and what it retains is now a live discovery issue, not an afterthought.

ABA Formal Opinion 512 remains the national baseline for lawyers' use of generative AI; the California State Bar approved updated generative-AI practice guidance on May 14, 2026.

These rulings turn on their specific facts and jurisdiction and don't establish a general rule. Counsel should confirm requirements with the court and client before relying on an AI-assisted workflow.

How it works

One workflow from legal hold to production

  1. 1

    Legal hold

    Issue holds and track acknowledgement across custodians, with a record of who was notified and when, before a single document is collected.

  2. 2

    AI-prioritised review

    Documents are scored for relevance and flagged for privilege and PII before a reviewer opens the queue, so attorneys spend time on what needs a legal judgment call.

  3. 3

    Court-ready output

    Redactions, privilege logs, and audit trails are built as review happens, not reconstructed afterward against a deadline.

  4. 4

    Production

    Generate production sets with Bates numbering and load files, with a full audit trail attached to every document that goes out.

AI under protective order

Built for review under a protective order

No AI model gets direct access to your systems. An internal model works inside your environment and strips sensitive material before anything goes out; the outside model sees OrcheSight as the caller, never your organization, and keeps nothing. The full log stays in your environment.

No direct system access

An outside AI model never touches your case data directly — it sees what OrcheSight passes it, and nothing else.

Nothing retained outside

The outside model keeps no copy of what it processed, so there is nothing for it to train on or disclose later.

The full log is yours

Every prompt and answer is recorded in your environment, so counsel can show exactly what an AI model saw and when.

Available as SaaS on Microsoft Azure, fully air-gapped on-premises for matters that can't touch the public internet, or as a managed service.

Beyond the legal department

One case record, not five department silos

A matter often touches HR, Cyber, and Compliance before it reaches Legal, evidence passing through spreadsheets, tickets, and email with no consistent chain of custody. OrcheSight keeps every department on one case record, role-based access included, so legal keeps a defensible account of how it got there.

Talk to us about your next matter

We'll walk through legal hold, AI-assisted review, and production on your infrastructure, and how the architecture holds up under a protective order.