Compliance & Legal

How to respond to a subpoena or legal hold with AI

Compliance & Legal3 AI tools7 steps6 friction points

Want an agent to handle this on your live data? See the Starch version →

Responding to a subpoena or legal hold means identifying every relevant document, message, email, and record your business holds — then preserving it, logging it, and often producing it in a specific format under a hard deadline. For most small operators, this isn't a routine process with a dedicated legal team running it. It lands on your plate without warning, and the stakes for getting it wrong — missing documents, improper handling, spoliation — are real.

The workflow feels like an AI problem because so much of it is language work: parsing dense legal language to understand scope, drafting litigation hold notices, summarizing what categories of documents are covered, and writing responses to counsel. These are exactly the kinds of tasks where a good LLM genuinely saves time — turning a page of legal boilerplate into a clear action list, or drafting a hold notice from a template, in minutes rather than hours.

ChatGPT, Claude, and Gemini can all meaningfully assist here. They're good at interpreting subpoena scope, drafting hold notices, writing custodian communications, and structuring your response checklist. What they can't do is reach into your actual email, Slack, or document systems — you bring the documents to them, not the other way around.

Skip the manual work → See the Starch version

Compliance & Legal3 AI tools7 steps6 friction points

AI walkthrough

How to do it with AI today

A practical walkthrough using ChatGPT, Claude, and other off-the-shelf LLMs — what they're good at, what you'll have to do by hand.

Tools that work for this

ClaudeChatGPTGemini

Step-by-step

1 Paste the full text of the subpoena or legal hold notice into Claude or ChatGPT and ask it to extract the scope: what time period is covered, what custodians are named, what categories of documents are requested, and what the response deadline is.

2 Use that extracted scope to build a collection checklist — ask the LLM to generate a line-by-line list of document types and systems you'll need to search (email, Slack, file storage, contracts, invoices, calendars).

3 Draft a litigation hold notice by pasting your company name, the relevant legal matter, the custodian list, and the document categories into the LLM, then prompting it to produce a compliant hold notice your team can send immediately.

4 For each system on your checklist (Gmail, Google Drive, Notion, etc.), manually search and export the relevant records. This is the step that cannot be automated with a raw LLM — you pull exports yourself and save them locally.

5 Paste individual documents or thread exports into the LLM to get plain-English summaries of what each contains and whether it falls within the stated scope of the subpoena.

6 Use the LLM to draft a privilege log scaffold — provide the list of documents and ask it to output a table with columns for date, author, recipient, subject, and privilege claim (attorney-client, work product, etc.).

7 Ask the LLM to draft your formal written response or objection letter based on the subpoena text and any specific objections your counsel has identified. Review and pass to your attorney before sending.

Prompts you can copy

Here is a subpoena I received. Summarize the document request scope in plain English: what categories of documents are covered, what custodians are named, and what the deadline is. [paste subpoena text]

Draft a litigation hold notice for a company called [Company Name] in connection with a lawsuit captioned [Case Name]. The hold covers all communications and documents from [date range] related to [subject matter]. Recipients are [names].

I need to identify which of the following document categories are likely responsive to this subpoena. For each category, tell me yes, no, or maybe, and explain why. Categories: [list]. Subpoena text: [paste].

Create a privilege log template for the following documents. For each, note date, sender, recipient, general subject, and whether it may be protected as attorney-client privileged or attorney work product. Documents: [paste list].

Draft a letter responding to opposing counsel acknowledging receipt of this subpoena, stating we are in the process of collecting responsive documents, and noting our anticipated production date of [date]. Tone should be formal and professional.

Reality check

Where this gets hard

The walkthrough above works — until your numbers change, the LLM hallucinates, or you have to re-paste everything next month.

Every session starts from scratch — the subpoena text, your custodian list, and your document inventory all have to be re-pasted each time you open a new conversation.

The LLM has no access to your actual email, Slack, Drive, or Notion — you manually export and paste content, which means your collection is only as complete as what you remember to pull.

Large document sets hit context limits fast. A Gmail export from one custodian over a two-year period can easily exceed what fits in a single session, forcing you to chunk and re-chunk manually.

Nothing is preserved between conversations — the privilege log, the hold notice, the checklist all live in a chat window. Close the tab and you're rebuilding the paper trail from memory.

Prompt drift is real for long projects. The structured output you carefully engineered in week one doesn't reliably reproduce in week three when you're following up or adding new custodians.

There is no audit trail. If opposing counsel or a court ever asks how you managed the legal hold process, a series of chat exports is a weak answer compared to a documented, timestamped workflow.

Tired of the friction?

Starch runs the whole workflow on live data — no copy-paste, no hallucinated numbers, no re-prompting next month.

See the Starch version →

Starch alternative

The same workflow on Starch

Starch is an agentic operating system — for this workflow, that means an agent builds a persistent legal hold tracker connected to your actual email, documents, and task systems, so the process runs against live data instead of manual exports.

Connect Gmail through Starch's scheduled sync — Starch indexes your actual messages on a schedule, so when a subpoena scope is entered, the agent can search your real inbox rather than waiting for you to paste exports manually.

Use Starch's Knowledge Management app to store the subpoena text, hold notice, custodian list, privilege log, and response drafts in one searchable place — so the full paper trail is organized and retrievable, not scattered across chat windows.

Use Starch's Task Manager to capture every action the hold requires — collect from these five custodians, send hold notice by this date, produce documents by this deadline — with P1 priority flags and overdue alerts so nothing slips.

Describe the tracker you need in plain English: 'Build me a legal hold log that tracks each custodian, their assigned document categories, collection status, and review notes.' The agent builds it as a persistent app your whole team can use.

Connect Google Drive or Notion from Starch's integration catalog — the agent queries those systems live when running a collection task, so your document inventory reflects what actually exists today, not a snapshot you pulled last week.

Because apps and automations persist in Starch, you can reopen the same legal hold workflow weeks later — add a custodian, update a deadline, check collection status — without re-prompting from scratch or reconstructing context.

Get closed-beta access →

Toolkit