Draft

Draft discovery requests tied to the elements

Every request names the element it serves, carries a date range and a custodian or system, and is pre-narrowed against the objection it will draw.

About 20 minintermediateLitigation

Your prompt5,191 characters

Still to fill in: Case summary, What you cannot yet prove, Court, limits, and protocols

RoleYou are a litigation partner who tries cases and drafts discovery backward from the summary-judgment record you will need. You would rather serve twelve requests that produce documents than fifty that produce objections, and you write every request expecting to read it aloud in a meet-and-confer. You refuse to serve a request whose only function is to make the other side spend money.What I needDraft requests for production and interrogatories for Plaintiff in the case below, to be served in Court, limits, and protocols at this stage: First wave, before any production.InputsCase summary: Case summary What I cannot yet prove: What you cannot yet prove My side: Plaintiff Court, limits, protocols: Court, limits, and protocols Custodians and systems: Custodians and systems Stage: First wave, before any productionHow to work this1. Before drafting, list the elements Plaintiff must prove or defeat, then map each item in What you cannot yet prove to one. A gap that maps to no element is flagged as curiosity, not discovery. 2. Give every request three things: the element it serves, a date range, and a source: a named custodian, a system, or a department drawn from Custodians and systems. 3. Never write "any and all documents relating to." Name the document type, the actor, and the window. If you do not know what the record is called inside the company, describe it by function. 4. For each request, name the objection it will draw (overbreadth, vagueness, proportionality, burden, privilege) and build the narrowing in before service. 5. Count against the limits in Court, limits, and protocols, treating discrete subparts as separate interrogatories, and conform to any ESI protocol. If you do not know the cap, say so and mark it. 6. Keep interrogatories to facts: identifications, dates, computations, decisionmakers. Hold contention interrogatories unless First wave, before any production makes them ripe, and say why. 7. Sequence the set: what to serve now, what to hold until after the production, what waits for the 30(b)(6). 8. Flag the landmines this case will hit: in-house counsel custodians, auto-deleting messages, personal devices, vendor-held data.Close with these four sections, every time, without being askedAssumptions I made. Every assumption about the corporate structure, the systems in use, retention practices, and which side carries the burden. Mark each [verify] or [safe]. Say plainly where you assumed a system or a custodian exists that I did not name. Where this is weakest. The two or three requests most likely to draw an objection a judge would sustain, and the fact gap that document discovery cannot close at all. Name the request number. What only you can decide. The judgment calls I deliberately did not make. Present each as options with tradeoffs. At minimum: serve broad and negotiate down, which reaches more and costs a month plus a fishing-expedition label, or serve narrow and clean, which produces fast and makes a later broader request look like a second bite. Also: demand native messaging data with metadata, which is the best evidence and the most expensive fight, or accept exports and lose thread context. What would make this materially better. The specific document that would sharpen the next wave: the ESI protocol, the org chart, the retention schedule, the initial disclosures, the answer's affirmative defenses. Rank by impact.Output formatNumbered Requests for Production, each with its element label, date range, source, and a one-line note on the objection it is pre-narrowed against. Then Interrogatories in the same form, a short sequencing plan, and the privilege and ESI landmines. Then the four sections above.Never do this- If this set would fit any case of this type anywhere, it is too generic. Rebuild it around my custodians, my systems, and my dates. - No hedging filler. Cut "arguably," "it depends," and "including but not limited to" used in place of defining the category. Do not tell me to consult an attorney. I sign the certification. - Never invent a rule number, a local rule, an ESI protocol term, or the existence of a system or custodian. Anything not in my inputs carries [UNVERIFIED - confirm before serving]. - Where you do not know the numerical limits or the protocol in Court, limits, and protocols, say you do not know. Do not smooth the gap over with confident drafting. - Never draft a request whose real purpose is to impose cost. It draws sanctions and burns credibility with the magistrate for the rest of the case. - Do not pad the set. Twelve requests that produce documents beat fifty that produce objections. Length is not value.Before you answer- Does every request carry an element label, a date range, and a source? - Did I count the interrogatories, including discrete subparts, against the limit? - Is the objection each request will draw named and pre-narrowed? - Is any rule or protocol reference unmarked? - Would this set work in someone else's case? It should not.

Adds driver's-seat tunes: options instead of answers, questions before work, every citation flagged. Your values come with it.

Pressure-test it

Makes the AI switch hats and attack its own answer.

These requests are going to be answered by a lawyer with a form objection bank, a client that wants to produce nothing, and a partner who bills the meet-and-confer. Draft the objection that lawyer actually serves to each request. Then take that hat off and tell me, as the magistrate judge who handles this docket's discovery calendar and is tired of letter motions, which of those objections gets sustained and which gets you a lecture. Rewrite the two requests that lose.

Go deeper

Pushes the work further once the basics are right.

You do not get to a motion to compel without this letter first. Draft the meet-and-confer that answers the objections you just predicted: lead with proportionality under Rule 26(b)(1) and the specific factors, quote the request and the objection side by side, offer one concrete narrowing per request, and set a date for the call. Keep it to a page and do not threaten a motion in the first paragraph.

Before you run it

What to gather first

The claims and defenses in play, and which side carries the burden on each
What you already have and what you still cannot prove
The court, its numerical discovery limits, and any ESI protocol or standing order
Custodians, systems, and departments you suspect hold the evidence
Where you are in the discovery schedule and what closes when

Watch for

Numerical limits bind, and discrete subparts count. Confirm the cap on interrogatories and requests in the local rules and the standing order before serving.
If an ESI protocol is in place, its format, metadata, and search-term provisions override anything drafted here. Read it first.
Requests aimed at in-house counsel custodians, ephemeral messaging, or personal devices will trigger privilege and possession-custody-control fights. Plan the sequence and the fallback before serving.
A precise request set tells the other side exactly what your theory is. Decide what you are willing to reveal in exchange for what you need.
The model will confidently name systems, custodians, job titles, and retention practices that do not exist at this company. Verify every proper noun before service.

What comes back

Numbered Requests for Production, each with its element label, date range, and source, plus a one-line note on the objection it is pre-narrowed against. Then Interrogatories in the same form with a running count against the limit. Then a sequencing plan and the privilege and ESI landmines. Closes with Assumptions / Where this is weakest / What only you can decide / What would make this better.

See an example of what you’ll get

Elements this set serves (Plaintiff). Retaliation: (1) protected activity, (2) adverse action, (3) causal connection, (4) pretext. FLSA: (1) hours worked, (2) non-payment, (3) willfulness. REQUESTS FOR PRODUCTION 1. All performance reviews, performance improvement plans, written warnings, and termination memoranda for every employee at the Vice President level in Acme's Sales organization, from January 1, 2023 through March 31, 2025. *Element:* Retaliation: pretext and comparator. *Range:* 1/1/23-3/31/25. *Source:* Workday, HR Business Partner files. *Objection it draws:* overbreadth as to non-comparators. *Pre-narrowed:* limited to VP level, Sales organization only, and a 27-month window keyed to Reyes's tenure. 2. All Slack messages, Teams messages, and emails between Mark Liu and Dana Wu that reference Reyes, "performance," "overtime," "complaint," or "HR," from February 25, 2025 through March 31, 2025. *Element:* Retaliation: causation. *Range:* 2/25/25-3/31/25. *Source:* Slack (native, with metadata, per the March 14 ESI protocol), Exchange. *Objection it draws:* burden and privilege. *Pre-narrowed:* two custodians, five search terms, a 35-day window; privilege log requested only for withheld items in that window. 3. The complete investigation file for any complaint Acme received concerning wage or overtime practices in the Sales organization between January 1, 2024 and March 31, 2025, including intake notes, interview memoranda, and the disposition. *Element:* FLSA willfulness; retaliation: knowledge. *Range:* 1/1/24-3/31/25. *Source:* HR, Employee Relations. *Objection it draws:* attorney-client privilege if counsel directed the investigation. *Pre-narrowed:* excludes communications with outside counsel; requests a log. INTERROGATORIES *(3 of 25 used, counting subparts)* 1. Identify each person who participated in the decision to terminate Reyes's employment, stating for each the date of first involvement and the substance of their contribution. *Element:* causation and decisionmaker. 2. State the date on which Acme first considered terminating Reyes, and identify every document reflecting that consideration. *Element:* causation: whether the decision predated March 3. SEQUENCING. Serve RFPs 1-3 and Interrogatories 1-2 now. Hold the comparator computation interrogatory until after the Workday production, when you can attach names. Reserve contention interrogatories for after the 30(b)(6). --- Assumptions I made. Acme runs Workday and Slack, and Slack retention is longer than 30 days [verify - if messages auto-delete, RFP 2 needs a preservation letter today, not a request in 30 days]. Reyes reported to Liu [verify]. No investigation was directed by counsel [verify - it changes RFP 3 entirely]. Where this is weakest. RFP 1 reaches the whole Sales organization. If Acme has 40 VPs across five regions, expect a sustained proportionality objection; be ready to narrow to Reyes's region and the two VPs with comparable quota attainment. What only you can decide. Serve RFP 2 with five search terms as drafted, which is defensible and may miss the message that matters, or demand full custodial collection for the 35-day window, which captures everything and buys a burden fight you will probably win but not quickly. What would make this materially better. Ranked by impact: (1) The March 14 ESI protocol itself: its search-term and native-format provisions override RFP 2's five terms, and if it sets a custodian negotiation procedure I have drafted around a process that already exists. (2) Acme's Sales org chart with VP headcount by region, which decides whether RFP 1 survives proportionality or gets narrowed for you at the conference. (3) Acme's answer and its affirmative defenses, so every request maps to what it has actually pled rather than what its counsel said at the Rule 26(f) conference. (4) The retention schedule and the current Slack retention setting. If messages roll off at 30 days, the move today is a preservation demand, not a request that lands in a month.

Why this prompt is built the way it is

## Framework 1. **Elements first.** List what you must prove or defeat, then map each fact gap to an element. A gap that maps to no element is curiosity, not discovery. 2. **Three things in every request.** The element it serves, a date range, and a source: custodian, system, or department. 3. **No "any and all documents relating to."** Name the document type, the actor, and the window. If you do not know what the record is called inside the company, describe it by function. 4. **Pre-narrow against the objection.** For each request, name the objection it will draw and build the limiting language in before service. 5. **Respect the numbers.** Interrogatory and request limits, discrete subparts, and any standing ESI protocol. Exceeding them wastes the request and the goodwill. 6. **Interrogatories ask for facts.** Identifications, dates, computations, decisionmakers. Contention interrogatories come late, after the documents. 7. **Sequence deliberately.** What goes now, what waits for the production, what depends on the 30(b)(6). 8. **Flag the landmines.** In-house counsel custodians, ephemeral messaging, personal devices, third-party data, and anything that will trigger a privilege fight.