Legal Hold Automation: Exterro vs Everlaw vs AI Agent

By Jude Lee · · Comparison

Small-firm legal team reviewing a litigation hold tracking list on a monitor in a conference room

Legal tech vendors are racing to put agents behind their products right now; LawFuel covered Casepoint’s agentic push as part of the broader move “beyond the chatbot.” For a five-lawyer litigation shop, the interesting question isn’t which vendor announced an agent. It’s which parts of your preservation workflow an agent should touch — and which parts it must never touch alone.

Why preservation is the workflow most worth automating first

Under Federal Rule of Civil Procedure 37(e), if electronically stored information that should have been preserved is lost because a party failed to take reasonable steps to preserve it, and it can’t be restored or replaced, a court may order measures no greater than necessary to cure the prejudice — and, only on finding the party acted with intent to deprive another party of the information, may presume the information was unfavorable, instruct the jury accordingly, or dismiss. Read the rule text itself before relying on that paraphrase — Rule 37 as published by Cornell’s Legal Information Institute — and check your state’s analog, because state practice varies and some jurisdictions are stricter.

The operative words are “reasonable steps.” Reasonableness is proven with a record: who you notified, when, what you told them, who acknowledged, who didn’t, what you did about the silence, and when you released the hold. That record is clerical work — which is what agents are good at, and what partners forget to do in month four of a quiet case.

Three ways small firms run holds today

Manual. A Word template, an email, a spreadsheet, and a calendar reminder someone eventually stops honoring. It works until you have more than a handful of active matters, or until opposing counsel asks how you tracked acknowledgments.

A dedicated hold module. Exterro and Everlaw both publish legal hold products, and both give you notice templates, acknowledgment tracking, automated reminders, custodian questionnaires, and an audit log designed to be shown to a court. That last part is the real product.

A custom agent on top of the systems you already run. Your matters, parties, and contacts already live in Clio, MyCase, Smokeball, or NetDocuments. An AI assistant connected to those systems through MCP — the open Model Context Protocol that gives an assistant governed access to specific tools and data — can read the matter, draft a hold notice tailored to the claims, send it, watch for replies, chase non-responders, and write the trail back to the matter file. We walk through the plumbing in connecting an AI assistant to your firm’s matter data with a custom MCP server.

How the dedicated products differ, and what to check before you buy

Exterro and Everlaw are not interchangeable, and the differences that matter to a small firm are structural rather than feature-list trivia. Broadly, Exterro positions legal hold within a wider legal GRC and e-discovery portfolio aimed at corporate legal departments, while Everlaw’s Legal Hold sits alongside its cloud-native e-discovery and review platform. Packaging and deployment change release to release, so treat the following as the questions to ask rather than answers:

Verify all of it in each vendor’s current product documentation. Pricing pages and feature matrices age fast, and anything a blog tells you about tiers is stale within a year.

Dedicated hold module
Defensibility features are built and tested in real disputes. Fast to stand up. The vendor maintains the audit log — and because your firm can’t edit it, it’s harder to impeach. Trade-offs, depending on tier and configuration: subscription cost recurs regardless of hold volume, the tool may sit apart from your practice-management system so custodians get entered twice, and notice drafting is generally template-driven rather than aware of the specific matter. Confirm each against the vendor’s own docs.
Custom hold agent over your own stack
Reads the actual matter (parties, claims, opposing counsel, key dates) so notices are specific rather than generic. Lives inside the system of record, so no double entry. You own the logic and can extend it to adjacent workflows. But you also build and maintain the defensibility — append-only logging, trustworthy timestamps, access controls no one at the firm can override — and that is precisely the product you chose not to buy.
If your hold system can’t produce a clean, timestamped record without a paralegal reconstructing it from email, you don’t have a hold system — you have a habit.

What a hold agent actually does, step by step

  1. Trigger on a real event, not a vibe

    Fire when a matter status changes to “litigation anticipated,” when a demand letter is logged, or when a complaint is filed. The trigger is a rule, not a judgment — the attorney sets the status.

  2. Assemble the custodian list from your own data

    The agent pulls related contacts from the matter, proposes a custodian list with its reasoning, and presents it for approval. It should surface who it could not classify rather than guessing.

  3. Draft a matter-specific notice from a firm skill

    A skill — a packaged, reusable instruction set that makes the assistant do the job the same way every time — encodes your firm’s hold language, required elements, tone, and the systems and data types to name explicitly.

  4. Send after human approval

    A lawyer reviews and approves. The agent sends, logs the send, and records delivery.

  5. Chase acknowledgments on a schedule

    Reminders at your chosen intervals, escalation to the client contact after N days of silence, and a standing list of who is outstanding.

  6. Re-issue periodically and on personnel changes

    Periodic reminders keep the duty alive; a new custodian triggers a fresh notice.

  7. Release the hold on instruction and archive the record

    Only on attorney instruction. The full chronology writes back to the matter as a single defensible document.

Most of that is scheduling and record-keeping rather than language generation. That’s the tell for a good agentic candidate.

Where these agents break

They invent custodians when the matter data is thin — a contact in Clio with no role field becomes a guess. They cannot see systems they aren’t connected to, so BYOD phones, personal email, and third-party vendors remain a human interview. They will send a technically perfect notice to the wrong entity if the matter record names the wrong party. And the immutability problem is the one to solve first: a log your own users can alter invites a challenge you don’t want to answer under oath.

An honest rule of thumb, stated as opinion: if you run a small number of preservation-heavy matters a year, buy the module or stay manual with a disciplined checklist. If holds are constant and your custodian data already lives in your practice-management system, a custom agent starts to pay. Either way, set the oversight model deliberately — see three oversight models for supervising AI agents.

Modeling the payback without making up numbers

Don’t take anyone’s dollar figure, including ours. Build your own worksheet.

Where preservation sits in a firm’s AI stack

As of 2026, most small and mid-size firms run three layers: a general assistant (Claude, ChatGPT, or Copilot) for drafting and analysis; AI features inside platforms they already pay for, such as Clio Duo or Smokeball Archie; and, in a growing minority, purpose-built legal tools like CoCounsel or Harvey, or a custom agent stitched onto their own stack. Document review, record summarization, first-draft correspondence, and intake qualification get the attention. Preservation tracking is still mostly manual — which is exactly why it’s available as an easy win.

Where is your firm losing billable hours?

Get a free automation audit: we map your intake-to-invoice workflow and show you exactly what's worth automating — before you spend a dollar.

Get a free automation audit

Related guides