Every sales team I've worked with spends more time writing than talking to prospects. Cold emails, follow-ups, discovery notes, deal summaries for the weekly pipeline review, the "just checking in" message you send fourteen times a day. It's a writing job with a quota attached.
These are the ten prompts I actually use. Each one follows the same shape — Role → Context → Constraints → Output — and each one was tested against the real thing: my last three quarters of outreach, discovery notes, and objection logs. Fill the bracketed fields, run it, edit the result. None of them are "write me a sales email." They're specific instruments for specific moments.
If you want all 30+ prompts I use across sales, marketing, and customer success in one file, the pack is here — but the ten below are free to copy.
The structure I use for every prompt
Before the prompts, the structure — because once you internalize it you can write your own.
- Role — who the model is pretending to be. Not "you are a salesperson." A specific one: "a B2B SaaS AE with 8 years selling to mid-market ops teams."
- Context — the situation. The prospect's industry, last quarter's results, what they said on the last call. Concrete, not vague.
- Constraints — the hard rules. Word count, banned phrases, required format, tone. This is where most prompts fail — too loose.
- Output — exactly what you want back. "Eight subject lines" is better than "ideas for a subject line."
Every prompt below fills those four slots. When you write your own, ask: did I tell it who to be, what's happening, what it can't do, and what shape the answer takes?
1. Cold outreach that earns a reply
The moment: You found a prospect. You need a first email that doesn't read like every other cold email.
Role: You are a B2B SDR who books 3+ meetings a week from cold email,
selling [product] to [buyer role] at [company size] companies.
Context: The prospect is [name], [title] at [company]. Their team
[recent trigger — funding round, hiring spree, exec change, product
launch, public complaint]. Our product helps with [specific problem].
Constraints:
- Subject line: 5 words or fewer, lowercase, no question marks,
no exclamation points, no "quick," no "checking in."
- Body: 90 words max. First sentence references the trigger.
Second sentence is the problem we solve — stated as pain,
not as feature. Third sentence is one proof point (a number,
a logo, a before/after). CTA is a question, not a meeting link.
- Banned: "I hope this finds you well," "revolutionary,"
"game-changing," "synergy," "circle back," any emoji.
Output: 1 subject line + 1 email body. Then 2 alternate subject
lines, each a different angle (urgency / curiosity / referral-style).
The alternates matter. Your first subject line is rarely your best one.
2. Discovery-call prep in 4 minutes
The moment: You have a discovery call in an hour. You need to walk in knowing what to ask.
Role: You are a senior AE preparing a discovery call.
Context: Prospect is [company], [industry], ~[employee count]
employees. They booked after [source: inbound demo request /
outbound reply / referral from X]. Public info: [paste 3-5 bullets
from their site, a recent press release, or their job postings].
Constraints:
- Identify the 3 most likely business problems we solve for
someone in their role. Rank by likelihood.
- For each: write 2 diagnostic questions that would confirm or
rule it out. Questions must be open-ended (no yes/no).
Each question should surface a number, a frequency, or a
workflow — not an opinion.
- End with 1 "landmine" question — something that, if they
answer it a certain way, kills the deal. State what the
deal-killing answer sounds like.
Output: A numbered list. No preamble. No "great question!" energy.
The landmine question is the point. Most discovery prep lists questions you want to ask. This one forces you to surface the question you don't want the answer to.
3. Objection handler that doesn't sound defensive
The moment: The prospect said "it's too expensive" / "we need to think about it" / "we already use [competitor]." You need a response that isn't a rebuttal script.
Role: You are a consultative AE. You do not "overcome" objections.
You diagnose them.
Context: Prospect said: "[exact objection, verbatim]."
What we know: [their budget cycle / their current solution /
their stated priority / who else is involved in the decision].
Constraints:
- First, name the 3 possible real meanings behind this objection.
"It's too expensive" almost never means the price is too high.
List what it actually might mean (no budget this quarter,
haven't seen value yet, comparing to a cheaper tool, needs
boss approval, using it as a polite no).
- For each possible real meaning: write the one question that
would tell you which one it is. No defending, no justifying,
no "but actually our ROI is…"
- Mark which real meaning is most dangerous (the one that kills
the deal no matter what you say) and what you'd do if it's that.
Output: The 3 meanings, the 3 diagnostic questions, and the
diagnosis. Plain text. No motivational framing.
This prompt exists because most objection-handling training teaches you to argue. Good sellers diagnose.
4. The "just checking in" replacement
The moment: You've been ghosted. You need to follow up without saying "just checking in."
Role: You are a persistent-but-not-annoying AE following up on
a stalled deal.
Context: Last contact was [date]. Last topic was [what we
discussed — their stated pain, the demo they saw, the question
they were going to ask their boss]. Since then: [anything new —
a feature shipped, a case study from their industry, a regulatory
change affecting them].
Constraints:
- 60 words or fewer.
- Do NOT use the phrase "just checking in," "circling back,"
"following up," "touching base," "wanted to see," or "bumping this."
- The email must contain exactly one new piece of value —
a fact, a stat, a relevant change. Not a "thought you'd find
this interesting" — a specific thing tied to their situation.
- CTA is either: (a) a question about their timeline, or
(b) a single binary choice ("worth a 15-min call this week,
or should I follow up next quarter?"). No open-ended "let me know."
Output: 1 email. Then 1 shorter alternate (under 35 words) that
leads with the value, not the context.
The short alternate usually wins. Long follow-ups signal you're anxious; short ones signal you're busy and they should be too.
5. Deal update your VP actually reads
The moment: It's pipeline review. You need to summarize a deal in a way that surfaces what your manager needs to know — not everything that happened.
Role: You are an AE writing a deal update for your VP.
Your VP reads 200 of these a week. They skim.
Context: Deal is [company], [stage — discovery/demo/proposal/
negotiation], [ACV], [target close date]. Last week's status was:
[copy last update]. This week: [the 2-3 things that happened —
call happened, doc sent, legal reviewed, champion changed].
Constraints:
- First line MUST be one of: "On track," "At risk," or "Slipped,"
followed by the one-sentence reason. No other first line allowed.
- Then 3 bullets max: what changed, what's next, what you need
(from the VP or anyone). Each bullet under 20 words.
- If "At risk" or "Slipped": the third bullet must be the specific
ask — "I need you to call [champion's boss] by Thursday" or
"I need legal to expedite redlines." Vague asks are banned.
- Word cap: 120 words total.
Output: The update. Plain text, no headers. Ready to paste into
CRM notes.
The "first line must be On track / At risk / Slipped" rule is the single highest-leverage constraint. It forces you to have a point of view before you write the rest.
6. Post-call notes that become the CRM record
The moment: Call just ended. You have 40 minutes of notes. You need them turned into something your team can use.
Role: You are an AE turning raw call notes into CRM-ready discovery
documentation.
Context: Raw notes (transcript, scratch notes, or memory dump):
[paste everything — messy is fine, abbreviations fine, half-thoughts
fine]. Deal context: [company, stage, what we're trying to sell].
Constraints:
- Extract and structure into exactly these sections:
1. Situation (2-3 sentences: who they are, why we're talking)
2. Pain points (list, each tied to a quote or paraphrase
from the notes — mark which is verbatim vs paraphrased)
3. Current state (what they do today, what tools, what process)
4. Decision process (who's involved, what timeline, what
criteria — only what was actually said, mark gaps as "UNKNOWN")
5. Next steps (with owner and date)
- Anything not said in the notes gets marked "not discussed,"
not inferred. Hallucinating a budget or a timeline from a vague
note is worse than leaving it blank.
- Mark anything that sounded like a commitment with ✅ and
anything that sounded like hesitation with ⚠️.
Output: The 5 sections. No editorializing. No "this is a great
opportunity" — let the facts imply that or not.
The "mark verbatim vs paraphrased" rule is what makes this safe to use. AI will confidently turn "they mentioned budget season is coming up" into "they have budget next quarter." Don't let it.
7. Competitive battle card from a messy win-loss log
The moment: You have six months of win-loss notes against a competitor. You need a one-page battle card.
Role: You are a sales enablement lead building a competitive
battle card from raw field notes.
Context: Competitor: [name]. Win-loss notes (last 6 months):
[paste your CRM notes, Gong summaries, rep recollections —
messy is fine]. Our product: [ours].
Constraints:
- Structure as:
1. Where we win (3 bullets max, each backed by a specific
deal or quote from the notes — no generalities)
2. Where we lose (3 bullets max, same standard of evidence)
3. The trap (the one thing reps keep doing that loses deals
against this competitor — name it bluntly)
4. The one-line pivot (the single sentence a rep should say
when the prospect brings up this competitor)
- Banned: "feature parity," "comparable offering," "robust
solution." Use specific capabilities or specific outcomes.
- If the notes don't support a claim, don't make it.
Mark as "insufficient data" instead.
Output: The 4 sections. One page. A rep should be able to
read it in 90 seconds before a call.
The "trap" section is the most valuable and the one AI is worst at producing on its own — it requires the model to find the pattern in what your reps keep doing wrong. That's why you feed it your actual notes, not a feature list.
8. Proposal that doesn't read like a proposal
The moment: Procurement asked for a written proposal. You don't want a 40-page document nobody reads.
Role: You are an AE writing a proposal a prospect will actually read.
Context: Prospect: [company], [industry], [team size]. Problem
we're solving: [their words, from discovery — not our marketing
copy]. What we're proposing: [product + scope + timeline].
Stakeholders involved: [list]. Their stated success criteria:
[what they said they'd measure].
Constraints:
- Structure:
1. The situation (3 sentences, in their language, describing
where they are today — no product names yet)
2. What success looks like (their criteria, restated so
they recognize them)
3. What we'd do (the scope — described as outcomes and
milestones, not features)
4. Investment (one line: price + term. No tiers, no "contact
us," no 3-column comparison)
5. What we need from them (the 2-3 things required to start)
- Maximum 2 pages. No executive summary, no table of contents,
no "thank you for the opportunity."
- Banned: "leverage" (as a verb), "best-in-class," "world-class,"
"seamless," "robust," "comprehensive."
Output: The proposal. Ready to send as a PDF.
The investment line is the constraint that scares people. One number. If your pricing genuinely needs three tiers, this prompt isn't for that deal — it's for the 80% of deals where a single clean number closes faster than a menu.
9. Referral request that doesn't feel transactional
The moment: A customer is happy. You want introductions without sounding like you're mining them.
Role: You are an AE asking a happy customer for referrals,
without making it feel like a transaction.
Context: Customer: [name] at [company]. We've been working
together since [date]. Recent win: [specific outcome they got —
a number, a solved problem, a moment]. Relationship: [warm /
very warm / just-delivered-big-win].
Constraints:
- Do NOT ask for "anyone who might benefit." That's a fishing
net and customers ignore it.
- Do NOT offer a referral fee or gift card unless that's genuinely
your company's policy — it changes the relationship.
- The email must propose 1-2 specific people or 1 specific
profile, based on what you know about their network/industry.
"Someone else running [their function] at a similar-stage
company" is specific. "Anyone in your network" is not.
- Make it easy to say no: explicitly state "no pressure, and if
no one comes to mind that's totally fine."
- Under 120 words.
Output: 1 email. Tone: peer-to-peer, not vendor-to-customer.
The reader should feel like you respect their time and their
relationships.
The specificity rule is everything. "Do you know anyone who needs this?" gets ignored. "I noticed you used to work at [their former employer] — is there anyone still there running ops who's dealing with [specific problem]?" gets a reply.
10. Quarterly business review that surfaces the renewal risk
The moment: It's QBR time. You need to present value delivered AND surface the things that might lose the renewal.
Role: You are a CSM/AE preparing a QBR for an existing customer.
Your job is two things: prove value, and surface renewal risk early.
Context: Customer: [company], contract value [ACV], renewal
date [date]. This quarter we delivered: [list]. Usage data:
[active users, frequency, features used, features not touched].
Known issues: [support tickets, complaints, churn signals].
Their original goals: [what they said they wanted at signing].
Constraints:
- Structure the QBR as:
1. Goals vs. reality (their stated goals at signing,
and where each one actually landed — on track, ahead,
or behind. Be honest. Hiding a miss loses the renewal
later, surfacing it now saves it.)
2. The one number (the single metric that best proves
value this quarter — with the before/after)
3. What we're worried about (the 1-2 risks to renewal —
low usage on a key feature, a champion who left,
a competitor being evaluated. Name it.)
4. Next quarter's bet (the one thing we'll focus on to
compound value before renewal)
- Banned: vanity metrics (logins, "engagement," total users
if it's not tied to outcomes).
- If a goal from signing is behind, say so and propose
a fix in the same sentence. Don't bury it.
Output: A 1-page QBR brief. Slide-ready, but text first.
The "what we're worried about" section is why this prompt exists. Most QBRs are victory laps. The renewals you lose are the ones where the risk was visible for months and nobody said it out loud.
How I tested these
Each prompt was run three times against three different real inputs (real prospects, real calls, real deals). A prompt that only works once is an anecdote, not a tool. The ones above held up across inputs.
Then I generalized: does the prompt work for a different product, a different industry, a different deal size? If a prompt only works for "SaaS sold to VPs of Engineering," it's not a prompt — it's a script. These ten generalized.
Finally, I read every output aloud. Anything that sounded like marketing copy got cut. Anything that sounded like a human who'd done the job got kept.
What's in the full pack
These ten are the sales-specific slice. The full productivity library has 30 prompts across sales, marketing, ops, and customer success — each tested the same way, each with the Role→Context→Constraints→Output structure broken out so you can adapt them. There's also a $9 launch pack if you want the core seven to start.
The point isn't the prompts. The point is the structure. Once you can write Role→Context→Constraints→Output for your own job, you stop needing other people's prompt libraries — including mine.
If you found this useful, the best thing you can do is steal the structure, write your own prompt for the worst part of your week, and test it three times. That's the whole method.
Top comments (0)