Discovery Call Prep Prompt

What this prompt does

A discovery call prep prompt turns research you paste — the prospect's website copy, a press release, their open job postings, the contact's LinkedIn summary — into a one-page brief for a single account: the pain points this company probably has, the questions that would confirm or kill each one, and the objections you'll hear in the last five minutes. It's account research, not a script.

That distinction matters because there's already a page here for the script. The sales call script prompt builds the reusable framework — opening, qualifying structure, close — that you carry from call to call. This prompt builds the part that changes every time: what you know about this company before you dial, and what you still need to find out.

The reason prep beats polish is that buyers stopped waiting for salespeople to diagnose them. Harvard Business Review called it the end of solution sales back in 2012 — Adamson, Dixon, and Toman's argument was that procurement teams now define their own solutions before a supplier ever hears from them. Walk into that call with "so tell me about your challenges" and you're asking someone to repeat homework they finished months ago.

One design rule runs through the whole prompt: the brief is built only from what you paste. Every pain point has to quote the line of research that suggests it, and anything the model infers beyond the evidence gets tagged [GUESS]. A confident brief you can't trace is worse than a thin one you can.

The prompt

Build me a pre-call brief for a sales discovery call. Work ONLY from the research I paste below plus what I tell you we sell. Do not use anything you believe you know about this company from your training data — it may be stale or belong to a similarly named company.

**What we sell:** [ONE LINE, PLUS THE 2-3 PROBLEMS IT SOLVES]
**The account:** [COMPANY NAME, ROUGH SIZE, INDUSTRY]
**Who's on the call:** [NAME, TITLE, ANYTHING I KNOW ABOUT THEIR ROLE]
**Call type:** [FIRST DISCOVERY / DEMO / RENEWAL]

**The research:**
[PASTE WHAT YOU HAVE: ABOUT-PAGE COPY, RECENT NEWS OR PRESS RELEASES, OPEN JOB POSTINGS, THE PROSPECT'S LINKEDIN SUMMARY, ANYTHING ELSE PUBLIC]

Output four sections, one page total:

1. LIKELY PAIN POINTS — at most 3, ranked. Each one must quote the specific line from my research that suggests it. If the research only supports one or two, give one or two and say so.
2. DISCOVERY QUESTIONS — for each pain point, one open question that would confirm it's real and one that would surface what it costs them.
3. LIKELY OBJECTIONS — the 2 most probable for this account, each with a one-line first response.
4. WHAT I'M MISSING — what the research doesn't tell me that I should find out on the call.

Rules:
- Tag anything that is inference rather than evidence with [GUESS].
- No generic questions. "What keeps you up at night" is banned — every question names something from their world.
- Keep it scannable in 60 seconds. This gets read between calls, not studied.

How to use it

1. Feed it the unglamorous sources. Job postings are the best research most reps never paste — a company hiring four ops coordinators in a month is telling you exactly what's breaking. Press releases say what leadership wants to be true; the careers page says what's actually on fire.
2. Read the quotes, not just the claims. Each pain point cites the research line it came from. If the quote doesn't really support the claim, cut the pain point before the call does it for you.
3. Rehearse the objections out loud. The brief gives you a first response, not a rebuttal script. The objection handling prompt is where you refine those responses once you've heard the same pushback three calls running.
4. Treat "what I'm missing" as the agenda. The gaps section is the actual call plan — those are the questions only the prospect can answer. For demos and QBRs, the meeting agenda prompt handles the structure around them.

Example output

Inputs: you sell a scheduling and payroll platform for hourly workforces. The account is Meridian Fulfillment, a 140-person third-party logistics company in Ohio. First discovery with their VP of Operations. Research pasted: their careers page, a local business journal piece on their new second warehouse, and the VP's LinkedIn summary.

1. Supervisor capacity is stretched by the second site. Careers page: "four warehouse supervisor openings, all posted within 30 days." The journal piece says the new building adds 60,000 square feet. Confirm: "Four supervisor roles in a month is a lot at once — is that the new building, or backfill?" Cost: "When a shift runs without a supervisor today, what breaks first?"

2. Scheduling across two sites is probably still manual. [GUESS] — nothing in the research names their current tools. Based on their size and the VP's LinkedIn line about "scaling ops without scaling headcount." Confirm: "How does a schedule get built today — who makes it, and in what?"

Likely objection: "We just spent this year's budget on the building." First response: agree the timing is real, then ask what the expansion does to labor cost per order if scheduling stays manual.

What I'm missing: no revenue signal, no named software stack, no sense of who owns payroll. All three are call questions, not research gaps you can close from your desk.

The honesty sits in the gap between those first two pain points: one is traceable to two quoted sources, the other admits it's a hypothesis. On the call, you present the first as an observation and the second as a question — and the [GUESS] tag is what tells you which is which.

Variations

Renewal or QBR

"This is a renewal call, not first discovery. The research I'm pasting is our own account history — usage notes, past tickets, last quarter's summary. Build the brief around what has changed since we last talked, and make every question about the delta, not the original sale."

When the research is thin

Small private companies sometimes have a one-page website and nothing else. Don't let the model pad. Add: "The research here is thin. Give me whatever the evidence genuinely supports — even one pain point — and put everything else in the missing section. A short honest brief beats a full invented one."

Batch prep for a call block

"I have six first-discovery calls today. I'll paste research for each account, labeled. Give me a three-bullet brief per account — top pain point with its quote, one confirming question, one likely objection. No full briefs."

Common pitfalls

• Don't ask the model what it knows about the company. Training data runs months behind and blends similarly named companies into one confident wrong answer. The paste-your-own-research rule isn't a formality; it's the whole design.
• Don't recite the pain points as facts on the call. They're hypotheses, and the discovery questions exist to test them. A rep who opens with a diagnosis they haven't verified sounds like a cold reader, and prospects can tell.
• Don't prep the brief and skip the follow-up. The twenty minutes after the call, while you still remember what they actually said, is when the follow-up email prompt earns its slot — fed with the real answers, not the pre-call guesses.
• Try instead: keep the brief open during the call and mark each pain point right or wrong as you go. A brief that was 2-for-3 teaches you what your research sources are worth; one you never scored teaches you nothing.

Who uses this prompt

• SDRs and AEs: ten minutes before every qualified first call, and the batch variation for prospecting blocks. More in prompts for sales reps.
• Freelancers and consultants: the client discovery call is the same job with a different title — the research is their site and your email thread.
• Founders selling their own product: the person with the least selling experience on the call and the most riding on it, which makes borrowed structure worth more.

ChatGPT, Claude, and Gemini all run this prompt unmodified. The two rules doing the real work — restrict the model to pasted documents, and let it say the evidence isn't there — come straight from Anthropic's guidance on reducing hallucinations, which recommends exactly this pairing: external knowledge restriction plus explicit permission to admit uncertainty. In a brief you'll skim at 2:58 before a 3:00 call, the [GUESS] tags are worth more than the guesses.