The Real Cost of Manual Hold Workflows at Scale
The standard playbook — assemble the custodian list, draft the hold notice, send individual emails, log responses in a spreadsheet — was designed for 20 custodians, not 120. At scale in federal litigation (particularly SDNY and NDCA matters where discovery obligations are enforced aggressively), the failure modes multiply: notices that go to stale email addresses, holdouts who never acknowledge and are never escalated, and a hold log in Relativity that reflects what you intended to send rather than what was actually confirmed. The pain isn't just inefficiency. A disputed or incomplete hold log in front of a federal magistrate is a sanctions risk. Firms absorbing $180K–$400K per engagement for litigation support cannot afford that exposure.
What an AI Agent Does Inside Relativity, Onna, and Nuix
An AI Labor Company preservation-hold agent connects to your existing Relativity environment and ingests the custodian manifest from case intake. From there, it issues hold notices automatically — across email and, where integrated, through iManage matter workspaces — and begins tracking acknowledgment status in real time. Custodians who haven't acknowledged by a configurable threshold get automated follow-up. Holdouts who reach the escalation window are flagged to the supervising partner with a pre-drafted escalation memo. Every action — issued, acknowledged, escalated, confirmed — is written to a signed hold log in Relativity. The whole cycle, from case intake to a defensible hold log, can typically complete within four hours. Teams running this workflow report 65–85% reductions in the manual time their staff spends on custodian outreach across a matter.
The Business Case: Capacity, Risk Reduction, and Billable Leverage
Preservation hold management is non-billable administrative work that consumes paralegal and litigation technology staff time on every new matter. An agent that compresses a two-day manual workflow into four automated hours isn't just faster — it frees your team to handle more matters simultaneously without adding headcount. At AmLaw 50 volumes, that capacity unlocks real revenue: each additional matter your staff can support without hiring adds directly to the firm's top line. The risk reduction side is equally concrete. A defensible, timestamped hold log in Relativity — produced automatically and consistently — removes one of the most common pressure points in early-stage discovery disputes. The agent can be live and producing results in approximately eight weeks from engagement start.
Does the agent modify anything in our Relativity environment without human approval?
The agent writes to the hold log and sends notices according to the workflow you configure, but escalation decisions — such as partner notifications and any matter-level judgments — require human approval before action. You define the thresholds; the agent executes within them.
What happens when a custodian's email bounces or they're on leave?
The agent flags undeliverable notices and no-response custodians at each escalation interval, routing them to the litigation technology team or supervising partner based on your configuration. Nothing falls through the cracks silently — every custodian status is tracked and auditable in Relativity.
We already have Onna and Nuix in our stack. Can the agent work with those?
Yes. The agent is designed to integrate with Onna for connector-based data source coverage and can surface relevant Nuix processing context alongside the hold workflow. The integration scope is scoped during the eight-week onboarding.