Most reps do not lose deals in the meeting. They lose them in the twenty minutes before it, when preparation gets compressed into a quick skim of the prospect's homepage and a glance at the CRM record.
Generative AI is unusually good at exactly this work. Pre-call prep is research, synthesis, and hypothesis generation — the things a language model does well — and it happens before anyone is on the line, so there is time to check the output before it matters.
Below are twenty prompts for ChatGPT or Claude, organized by the order you would actually use them: account research, stakeholder mapping, discovery planning, competitive preparation, and the first five minutes of the call. Each one is written to be copied as-is, with the bracketed parts replaced.
How to get materially better output from these
Four habits separate prompts that produce something useful from prompts that produce filler.
Give the model raw material, not just a company name. Paste the prospect's careers page, their last earnings call summary, the LinkedIn "About" section, your notes from the last call. A model reasoning over documents you supply is far more reliable than one recalling what it may have seen in training.
Say who you are and what you sell. Every prompt below assumes you will fill in your product and market. Without that, the output is generic sales advice rather than preparation for your specific conversation.
Ask for the reasoning, not just the answer. "Give me three hypotheses and the evidence for each" is more useful than "tell me their pain points," because you can evaluate evidence and you cannot evaluate an assertion.
Treat every factual claim as unverified. A model will produce a confident, specific, wrong number about a company's headcount or funding round. Use it to form questions, never to state facts to a prospect.
Account research (prompts 1–5)
1. Build a company brief from source material
You are a B2B sales researcher. I am selling [PRODUCT] to [ICP / TITLE].
Below is source material about [COMPANY]: their website copy, recent news,
and job postings.
<source>
[PASTE MATERIAL]
</source>
Produce:
1. What this company does, in two sentences a non-expert would understand
2. Three things that appear to have changed for them in the last 12 months
3. Two operational problems the material implies but does not state outright
4. Anything in the material that suggests budget, urgency, or a mandate
For each item, quote the specific line that supports it. If the material
does not support a claim, say so rather than inferring.
Why the quote requirement matters: it makes hallucination visible. If the model cannot point to a line, you know the claim came from somewhere other than your source.
2. Read the job postings for buying signals
Here are [COMPANY]'s current open roles:
<roles>
[PASTE JOB TITLES AND DESCRIPTIONS]
</roles>
I sell [PRODUCT], which helps with [PROBLEM].
What do these postings tell me about:
- Which teams are growing and which are not
- What tooling or process gaps the descriptions imply
- Which named role would most likely own a decision about [PROBLEM]
- Whether the language suggests they are building in-house instead of buying
Flag any posting that suggests they are already solving this internally.
Hiring is one of the few genuinely public, genuinely current signals about what a company is investing in. The last question is the important one — it surfaces the "we're building it ourselves" objection before the call rather than during it.
3. Turn an earnings call or annual report into sales context
Below is an excerpt from [COMPANY]'s most recent [EARNINGS CALL / ANNUAL
REPORT / SHAREHOLDER LETTER].
<document>
[PASTE]
</document>
I sell [PRODUCT] to [DEPARTMENT].
Extract:
1. The strategic priorities leadership named, in their own words
2. Any metric they said they are trying to move
3. Pressures they acknowledged (cost, competition, regulation, growth)
4. How [PRODUCT] could plausibly connect to one of the priorities above —
and be honest if the connection is weak
Quote directly. Do not paraphrase leadership's language; I want to use
their exact words on the call.
Using a buyer's own strategic language back to them is the cheapest credibility you will ever get. It also forces you to find a real connection rather than a generic value proposition.
4. Find the gap between positioning and reality
Here is how [COMPANY] describes itself publicly:
<positioning>
[PASTE MARKETING COPY]
</positioning>
And here is what customers/employees actually say:
<reviews>
[PASTE G2, GLASSDOOR, OR SUPPORT-FORUM EXCERPTS]
</reviews>
Where do these two diverge? For each gap, suggest one open question I
could ask that would let them describe the gap themselves without feeling
attacked.
The gap between how a company markets itself and how it actually operates is usually where the pain lives. The phrasing constraint matters — the goal is a question they want to answer, not a gotcha.
5. Prepare for an account you know almost nothing about
I have a call with [TITLE] at [COMPANY] in [INDUSTRY], roughly [SIZE]
employees. I sell [PRODUCT].
I have no other information.
Give me:
- Five things that are usually true of [TITLE] at a company this size in
this industry
- Three of those that would be most likely to be false here, and what
would make them false
- Four questions that would let me find out quickly which situation I am in
Mark clearly which parts are general pattern and which are specific
claims about this company. I expect there to be no specific claims.
This is the honest version of "research the account when you have no time." It produces hypotheses to test rather than facts to assume, and the final instruction keeps the model from inventing specifics.
Stakeholder mapping (prompts 6–9)
6. Infer the buying committee
I am selling [PRODUCT] (average deal size [AMOUNT], implementation touches
[TEAMS]) into [COMPANY], a [SIZE] company in [INDUSTRY].
My champion is [TITLE].
Map the likely buying committee: economic buyer, technical evaluator,
end users, and anyone with veto power (security, legal, procurement,
works council).
For each role, tell me:
- What they are measured on
- What would make them say no
- One question I should ask my champion to find out whether this person
is engaged
7. Prepare for a specific person
Here is [NAME]'s LinkedIn profile and any public writing or talks:
<profile>
[PASTE]
</profile>
They are [TITLE] at [COMPANY]. I sell [PRODUCT].
Based only on what is in this material:
- What has their career optimized for?
- What vocabulary and framing would land with them?
- What would they likely be skeptical of in a pitch for [PRODUCT]?
- What would a win look like for them personally in the next six months?
Note where you are inferring rather than reading.
8. Write the champion enablement brief
My champion is [NAME], [TITLE] at [COMPANY]. They have to sell [PRODUCT]
internally to [ECONOMIC BUYER TITLE].
Write a one-page internal brief my champion could forward, in their voice,
not mine. It must:
- Lead with the business problem, not the product
- Include the three numbers a CFO would ask about
- Pre-empt the two most likely objections
- Be under 400 words
Then list what I need to learn from my champion before this brief is
accurate rather than generic.
The second half is the valuable part. It converts an artifact into a discovery checklist.
9. Plan for the person who is not in the room
I have a call with [TITLE]. The person who actually controls budget is
[OTHER TITLE], who will not be there.
Help me:
- Identify what [OTHER TITLE] cares about that [TITLE] may not mention
- Draft two questions that surface how decisions like this actually get
made here
- Suggest a natural way to ask for [OTHER TITLE] to join the next call
without implying my contact lacks authority
Discovery planning (prompts 10–14)
10. Build the question set
I sell [PRODUCT], which helps [ICP] with [PROBLEM]. Discovery call with
[TITLE] at [COMPANY].
Write 12 discovery questions in this order: situation, then problem, then
impact, then decision process.
Rules:
- Open-ended only, no yes/no questions
- No question that presupposes they have the problem I sell against
- Include three that would help me disqualify quickly
- For each, add the follow-up I should ask if the answer is vague
The disqualification requirement is what makes this useful. Discovery that can only confirm interest is not discovery.
11. Pressure-test your own hypothesis
My hypothesis: [COMPANY] is struggling with [PROBLEM] because of
[REASON], and [PRODUCT] would help by [MECHANISM].
Argue against me. Give me:
- Three reasons this hypothesis might be wrong
- What I would observe on the call if it were wrong
- The single question most likely to disconfirm it quickly
Do not be diplomatic.
12. Plan the impact conversation
Prospect problem: [PROBLEM]. Their role: [TITLE]. Company size: [SIZE].
Help me quantify the cost of this problem with them, on the call.
Give me:
- The variables we would need (time, headcount, rates, error frequency)
- A simple arithmetic structure a buyer could follow out loud
- Three questions that get them to supply each number themselves
- What to do if they say they do not know
Do not invent benchmark figures. Leave every number as a blank for them
to fill.
The final instruction is the whole point. A number the buyer says is evidence; a number you say is a claim.
13. Design the demo around what you heard
Here are my notes from discovery:
<notes>
[PASTE]
</notes>
I am demoing [PRODUCT] next. Build a demo flow that:
- Opens by replaying their words back to them
- Shows only capabilities that map to something they actually said
- Names explicitly what I am NOT showing and why
- Ends at a decision, not a feature
List which of my product's capabilities should stay out of this demo.
14. Prepare for a technical evaluator
I have a call with [TITLE] who will evaluate [PRODUCT] technically.
Relevant facts about how it works: [ARCHITECTURE, DATA HANDLING,
INTEGRATIONS, SECURITY POSTURE].
Predict the ten hardest questions they will ask, ordered by likelihood.
For each: what a credible answer sounds like, and what an evasive answer
sounds like. Flag any question where the honest answer is a limitation,
and suggest how to state it plainly.
Competitive and objection prep (prompts 15–18)
15. Build a differentiation brief
I sell [PRODUCT]. The prospect is also evaluating [COMPETITOR].
Here is what I know to be true about both:
<facts>
[PASTE YOUR OWN COMPARISON NOTES]
</facts>
Write:
- Three differences that would matter to [TITLE] specifically
- Two areas where [COMPETITOR] is genuinely stronger, and how to
acknowledge them without conceding the deal
- One question that helps the buyer discover the difference themselves
Use only the facts I provided. Do not add claims about [COMPETITOR].
The constraint is essential. A model asked about a competitor from memory will produce plausible, outdated, sometimes defamatory claims. Supply the facts yourself.
16. Rehearse the objections
I sell [PRODUCT] at [PRICE POINT] to [TITLE].
List the eight objections I am most likely to hear, ordered by frequency.
For each:
- What the objection usually means underneath
- A clarifying question to ask before answering
- A two-sentence response
- The response that would make it worse
Then pick the one rep most often fumble and explain why.
17. Prepare for the status-quo objection
The prospect currently handles [PROBLEM] using [CURRENT APPROACH:
spreadsheet, manual process, incumbent tool, nothing].
Help me understand:
- Why that approach is genuinely reasonable for them today
- What would have to change for it to stop being reasonable
- Three questions that help them notice the cost they have normalized
- When I should conclude that doing nothing is the right answer for them
and disqualify
Do not treat the status quo as irrational.
Most deals are lost to no-decision, not to competitors. Preparation that respects the status quo produces better questions than preparation that dismisses it.
18. Stress-test your pricing conversation
My pricing: [STRUCTURE]. Prospect: [TITLE] at [SIZE] company.
Predict:
- How they will react to the number and why
- Which part of the structure will cause the most friction
- Three questions they will ask that I should have a crisp answer for
- What discount request is most likely, and what I should ask for in return
Give me the exact words for holding price without sounding rigid.
The first five minutes (prompts 19–20)
19. Write the opening
Call with [TITLE] at [COMPANY]. Context: [HOW THE MEETING CAME ABOUT].
My hypothesis about their situation: [HYPOTHESIS].
Write a 45-second opening that:
- States why I asked for the time, specifically
- Names my hypothesis and invites them to correct it
- Sets an agenda and asks if it is right
- Contains no pleasantries about the weather or their logo
Then give me a shorter version for when they join late and distracted.
20. Build the one-page call sheet
Using everything above about [COMPANY], produce a single-page call sheet:
- Three facts I must not get wrong
- My hypothesis, in one sentence
- Six questions in priority order
- Two objections and my one-line answers
- The specific outcome I want from this call
- The next step I will propose, and my fallback
Keep it under 250 words. I will read this in the 60 seconds before
joining.
If you use only one prompt from this article, use this one. It forces the rest of your preparation into something you can actually hold in your head while a person is talking to you.
Where AI preparation stops helping
Preparation has a hard limit: it all happens before the conversation, and the conversation rarely follows the plan.
The moment a prospect says something you did not anticipate — names a competitor you did not prepare for, raises a security requirement you have not seen, describes a problem in language your call sheet does not contain — your preparation is behind you and you are on your own. That is precisely the moment that decides the deal, and it is the moment no amount of pre-call prompting can reach.
This is the gap MagicScreen was built for: real-time intelligence during the call itself, on Zoom, Google Meet, and Microsoft Teams, without joining as a bot or recording the conversation. Prepare with the prompts above, and have something in the room with you when preparation runs out.
Download MagicScreen for Mac — free to start.