20 Claude Prompts for Sales (2026)

<context>
You are an enterprise account researcher supporting an account executive who sells [YOUR PRODUCT] to [TARGET INDUSTRY] companies. The AE is opening [PROSPECT COMPANY] as a target account this week.
</context>

<documents>
[PASTE 10-K SECTIONS, ANNUAL REPORT, OR EARNINGS CALL EXCERPT]
</documents>

<task>
Read the documents and produce an account brief the AE can absorb in five minutes. Extract: (1) stated strategic initiatives, quoted directly; (2) risks the company itself names that [YOUR PRODUCT] is relevant to; (3) financial signals that affect buying appetite — growth, margin pressure, announced cost programs; (4) leadership or org changes; (5) the single strongest why-now angle for outreach.
</task>

<constraints>
Every claim must quote or closely paraphrase the documents — no outside knowledge about the company. If a section has no support in the source, write "not in source" rather than filling the gap. Plain language, no marketing adjectives.
</constraints>

<output_format>
Five numbered sections matching the task list, 60 words maximum each, followed by a two-sentence "open with this" note for the first call.
</output_format>

<context>
You handle territory planning for a rep selling [YOUR PRODUCT]. The ideal customer profile: [ICP DESCRIPTION — SIZE, INDUSTRY, STACK, TRIGGER SIGNALS].
</context>

<accounts>
[PASTE YOUR ACCOUNT LIST, ONE PER LINE, WITH ANY NOTES YOU HAVE]
</accounts>

<task>
Score every account 1–5 against the profile using only the evidence in the list, then sort into three tiers: work now, nurture, ignore this quarter. For each account, name the single deciding piece of evidence and the single missing data point that would most change the score.
</task>

<constraints>
Never invent firmographic facts. An account with thin notes gets "unknown" on the relevant criterion, not a guess. If the profile description is too vague to score against, ask me up to 3 questions before scoring anything.
</constraints>

<output_format>
A table: account · score · tier · deciding evidence · missing data point. Below it, three sentences on the pattern you see in the work-now tier so the rep can hunt lookalikes.
</output_format>

<context>
An account executive is working a [YOUR PRODUCT] opportunity at [PROSPECT COMPANY], currently at [DEAL STAGE]. The goal is a map of who actually decides.
</context>

<people>
[PASTE NAMES, TITLES, AND YOUR NOTES ON EVERYONE THE DEAL HAS TOUCHED]
</people>

<task>
Map the buying committee: probable economic buyer, champion, technical evaluator, likely blocker, and any role that is simply missing from the list. For each person: what their role is measured on, what they personally risk if this purchase goes wrong, and one question the AE can ask to test your read.
</task>

<constraints>
Reason only from titles and notes; label anything beyond them as "inference". No personality profiling. If two people could plausibly hold the same role, present both with the evidence split rather than forcing a pick.
</constraints>

<output_format>
One block per person — role call, measured on, personal risk, test question — then a closing line naming the most dangerous gap in committee coverage.
</output_format>

<context>
A rep sells [YOUR PRODUCT]. Fit is strongest when these trigger conditions appear: [TRIGGER CONDITIONS — E.G., NEW LEADERSHIP, FUNDING, TOOL CONSOLIDATION, HIRING SPIKE].
</context>

<sources>
[PASTE PRESS RELEASES, NEWS EXCERPTS, AND JOB POSTS — LABEL EACH WITH ITS DATE]
</sources>

<task>
Identify every event in the sources that qualifies as a buying trigger for [YOUR PRODUCT]. Rank by strength and freshness. For the top two, write a one-sentence outreach angle that references the event without recycling its headline wording.
</task>

<constraints>
An event only counts if a source supports it — no inferred momentum. Flag anything older than 90 days as stale and rank it accordingly. The angle sentences must read like a person noticed something, not like a press-release summary.
</constraints>

<output_format>
Ranked list: event · source date · strength (strong / moderate / weak) with one-line reasoning. Then the two angle sentences, labeled with their target event.
</output_format>

<context>
You support a rep selling [YOUR PRODUCT] against [COMPETITOR LIST]. Target: [PROSPECT COMPANY]. The question is who already owns this budget line.
</context>

<evidence>
[PASTE JOB POSTINGS, TECH-STACK NOTES, CASE STUDIES, INTEGRATION PAGES — ANYTHING HINTING AT CURRENT TOOLING]
</evidence>

<task>
Infer the most likely incumbent vendor with a confidence rating (low / medium / high) tied to specific lines of evidence. List renewal-risk signals if any appear. Then build the displacement map: the three switching costs the prospect will raise, and an honest counter for each — honest meaning it concedes what is genuinely hard about switching.
</task>

<constraints>
Every confidence rating must cite the exact evidence line. If the evidence cannot distinguish between two vendors, say so plainly. Never claim the incumbent is failing without support in the evidence.
</constraints>

<output_format>
Verdict line with confidence, an evidence table, then the three switching objections with counters of 40 words or fewer each.
</output_format>

<context>
An account executive sells [YOUR PRODUCT]. Recipient: [NAME], [TITLE] at [PROSPECT COMPANY]. Research on hand: [PASTE YOUR ACCOUNT BRIEF OR THE THREE STRONGEST FACTS YOU HAVE].
</context>

<task>
Draft a first-touch email built to earn the reply, not the delete: one specific observation from the research, one sentence of relevance connecting it to what [YOUR PRODUCT] does, one low-friction ask — [THE ASK — E.G., 15 MINUTES, A REPLY, A LOOK AT A TWO-LINE IDEA].
</task>

<constraints>
90 words maximum including greeting and sign-off. No marketing adjectives, no "I hope this finds you well", no compliment openers, no exclamation marks. Subject line of six words or fewer, lowercase, no bait. If the research is too thin to write a specific observation, ask me up to 3 questions instead of writing filler.
</constraints>

<output_format>
Subject line, then the email body, then one line naming which research fact is carrying the message.
</output_format>

<context>
A rep is building a three-touch email sequence over [SEQUENCE LENGTH IN DAYS] days to [PERSONA] at [PROSPECT COMPANY]. Product: [YOUR PRODUCT]. Research: [KEY FACTS ABOUT THE ACCOUNT].
</context>

<task>
Write three self-contained emails, each carrying a different angle: touch one opens with a problem observation; touch two explains the mechanism — how [YOUR PRODUCT] attacks that problem, concretely; touch three is a clean, unembarrassed close-out that makes "no" easy to say. No touch may reference the silence before it.
</task>

<constraints>
80 words maximum per touch. Banned phrases: "just following up", "bumping this", "circling back", and any subject line containing "quick question". Each touch must use a different research fact — if the research only supports one, say so and stop.
</constraints>

<output_format>
Three blocks — send day, subject, body — then one line on what to change if [PERSONA] keeps opening but never replies.
</output_format>

<context>
A rep wants to connect with [NAME], [TITLE] at [PROSPECT COMPANY] on LinkedIn because [REASON — A SPECIFIC OBSERVATION, POST, OR SHARED CONTEXT]. The long game is a conversation about [YOUR PRODUCT]; the short game is simply being worth accepting.
</context>

<task>
Write two messages: (a) a connection note of 280 characters or fewer that gives a real reason to accept and pitches nothing; (b) a follow-up for after acceptance, 90 words or fewer, that opens the problem space and earns the right to keep talking — without asking for a meeting yet.
</task>

<constraints>
The connection note never mentions the product. The follow-up may name the problem space once and [YOUR PRODUCT] at most once. No flattery openers, no "great to connect", no fake curiosity about their content.
</constraints>

<output_format>
The note with its character count, the follow-up message, and one line on how long to wait between them and why.
</output_format>

<context>
A prospect replied to outreach about [YOUR PRODUCT] — but the reply is lukewarm: noncommittal, deflecting, or a soft "maybe later". The rep's goal: [YOUR GOAL — E.G., BOOK THE MEETING, GET REFERRED TO THE RIGHT PERSON].
</context>

<thread>
[PASTE THE FULL EMAIL THREAD, NEWEST MESSAGE LAST]
</thread>

<task>
First, diagnose in one sentence what the reply actually is — real interest, polite no, wrong person, or wrong timing — quoting the words that tell you. Then draft the response that matches the diagnosis: wrong-timing gets a date-anchored ask, wrong-person asks for the handoff and names the likely title, polite-no gets a graceful close that leaves the door open.
</task>

<constraints>
70 words maximum for the reply. Thank them at most once. Do not re-pitch the original email. Mirror their formality level.
</constraints>

<output_format>
Diagnosis line with quote, the reply draft, then the fallback move if this message gets silence.
</output_format>

<context>
[ATTENDEE NAME], [TITLE] at [PROSPECT COMPANY], attended [EVENT OR WEBINAR NAME] on [TOPIC]. What they actually did there: [ENGAGEMENT SIGNAL — QUESTION ASKED, POLL ANSWER, STAYED FOR THE DEMO]. Product: [YOUR PRODUCT].
</context>

<task>
Draft a follow-up email that starts from what they did, not from the fact that they attended. Connect their specific signal to one concrete next step: [THE OFFER — E.G., THE SLIDE THEY ASKED ABOUT, A 15-MINUTE WALKTHROUGH OF THAT ONE TOPIC].
</task>

<constraints>
100 words maximum. Never open with "Thanks for attending". Reference session content only in ways a real attendee would recognize — no generic "great discussion" language. If the engagement signal field is empty, ask me for it instead of writing a generic follow-up.
</constraints>

<output_format>
Subject line and body, then a one-sentence variant of the opener for someone who registered but did not show.
</output_format>

<context>
An account executive just finished a discovery call about [YOUR PRODUCT] with [PROSPECT COMPANY]. Methodology, if any: [METHODOLOGY — OR "NONE"].
</context>

<transcript>
[PASTE THE FULL CALL TRANSCRIPT]
</transcript>

<task>
Debrief the call: (1) every pain the prospect stated, in their own words, quoted verbatim; (2) every number they mentioned — team size, budget signals, timelines, current costs; (3) every commitment made by either side; (4) the discovery questions that were never asked or never answered — judged against what a complete discovery should establish: cost of the problem, decision process, timeline, and money; (5) a recap email ready to send.
</task>

<constraints>
Quotes must be verbatim from the transcript — no cleaned-up paraphrases presented as quotes. The recap email is 120 words maximum with commitments as bullets. If the transcript shows no discussion of money at all, flag that first, in bold.
</constraints>

<output_format>
Sections 1–4 as labeled lists, then the recap email in full.
</output_format>

<context>
A rep is reviewing the moment resistance showed up in a call about [YOUR PRODUCT]. They want to know what the objection actually was — not what it sounded like.
</context>

<transcript>
[PASTE THE TRANSCRIPT SEGMENT AROUND THE OBJECTION — A FEW MINUTES EITHER SIDE]
</transcript>

<task>
For every objection in the segment: classify the surface type (price, timing, authority, need, trust); hypothesize what sits underneath it, citing the exact wording that suggests your read; grade how the rep answered in the moment; then script a stronger response of 50 words or fewer that addresses the underlying issue rather than the surface words.
</task>

<constraints>
Never invent objections that are not in the segment. If the rep's live answer was good, say so and explain why — do not manufacture criticism. Scripted responses must be in spoken register: contractions allowed, no email formality.
</constraints>

<output_format>
Per objection: type · what is underneath · evidence quote · rep grade with a one-line reason · the better response.
</output_format>

<context>
Grade deal [DEAL NAME] at [PROSPECT COMPANY] for [YOUR PRODUCT] against MEDDPICC before the rep's manager does.
</context>

<deal_record>
[PASTE CALL NOTES, TRANSCRIPT EXCERPTS, AND EMAIL THREADS — EVERYTHING YOU HAVE]
</deal_record>

<task>
Grade each letter — Metrics, Economic buyer, Decision criteria, Decision process, Paper process, Identified pain, Champion, Competition — as evidenced, weak, or absent. An "evidenced" grade must cite a line from the record. For every weak or absent letter, write the single question most likely to fill the gap on the next call.
</task>

<constraints>
Evidence means the record says it, not that it is plausible. Champion requires proof they have sold internally without the rep in the room — access and enthusiasm alone grade as weak. Do not average the grades into a deal score; the gaps are the point.
</constraints>

<output_format>
An eight-row table: letter · grade · evidence or gap question. Then the three gaps to attack first, ordered by deal risk.
</output_format>

<context>
Deal: [DEAL NAME], stage [DEAL STAGE], forecast category [CATEGORY — COMMIT / BEST CASE / PIPELINE], close date [CLOSE DATE], value [DEAL VALUE]. The forecast call is in [TIME UNTIL THE CALL].
</context>

<deal_record>
[PASTE RECENT NOTES, TRANSCRIPT EXCERPTS, AND EMAIL THREADS]
</deal_record>

<task>
Build the defense file: (1) the deal story in four sentences a sales leader would accept without interrupting; (2) the evidence that supports the current forecast category; (3) the three hardest questions leadership will ask, written the way they would actually ask them, each with an honest answer drawn from the record; (4) the one risk the rep should volunteer before anyone finds it.
</task>

<constraints>
Honesty beats polish: if the record does not support the category, say which category it does support and why. No hedging filler — "hopefully", "should be fine", and "tracking well" are banned. Every answer in section 3 cites the record or admits the gap.
</constraints>

<output_format>
Four numbered sections; section 3 as question-and-answer pairs.
</output_format>

<context>
Test whether [CONTACT NAME], [TITLE] at [PROSPECT COMPANY] is a real champion for the [YOUR PRODUCT] deal — or just a fan.
</context>

<evidence>
[PASTE THEIR EMAILS, MEETING BEHAVIOR NOTES, AND ANYTHING THEY HAVE DONE UNPROMPTED]
</evidence>

<task>
Score the contact against five champion behaviors: spends political capital on the deal, sells it when the rep is not in the room, shares inside information unprompted, pushes the process forward without being asked, and stands to win personally if it closes. Cite evidence for each behavior or mark it unproven. Then give a verdict — champion, fan, or cannot tell — and prescribe one test: a small ask a champion would complete this week and a fan would quietly dodge.
</task>

<constraints>
Enthusiasm is not evidence. Meeting attendance is not evidence. Forwarding a deck is weak evidence at best — say why if you count it.
</constraints>

<output_format>
Behavior checklist with evidence, the verdict with one-line reasoning, and the test ask written word for word.
</output_format>

<context>
Your team is responding to an RFP from [PROSPECT COMPANY] for [YOUR PRODUCT CATEGORY]. Every answer must come from approved facts — nothing improvised.
</context>

<rfp_document>
[PASTE THE FULL RFP — EVERY SECTION, EVERY QUESTION]
</rfp_document>

<company_facts>
[PASTE YOUR APPROVED FACT BASE — CAPABILITIES, CERTIFICATIONS, LIMITS, DIFFERENTIATORS]
</company_facts>

<task>
Extract every question and requirement from the RFP and number them. Draft an answer for each using only what appears in company_facts. Classify every answer: full, partial, or cannot answer from facts. For partials, state exactly what is missing. Never soften a gap into marketing language.
</task>

<constraints>
No invented capabilities, certifications, or numbers — an invented answer in an RFP response can lose the deal and worse; treat it that way. Answers of 120 words or fewer unless the question explicitly demands detail. Plain declarative sentences.
</constraints>

<output_format>
Numbered question-and-answer pairs with classification tags, then a summary: counts per classification and the cannot-answer list formatted for routing to subject-matter experts.
</output_format>

<context>
A deal review for [DEAL NAME] at [PROSPECT COMPANY]. The deal has run for [TIME PERIOD] across multiple calls, and memory has gone soft.
</context>

<transcripts>
[PASTE ALL CALL TRANSCRIPTS IN ORDER, EACH LABELED WITH ITS DATE]
</transcripts>

<task>
Reconstruct the deal from the full record: (1) a dated timeline of every commitment either side made, with who made it; (2) the status of each — kept, broken, or still open; (3) contradictions — every place the prospect's story changed between calls, quoting both versions side by side; (4) a momentum read: is their language warming or cooling across the calls, with examples of each.
</task>

<constraints>
Every entry cites its call date. A contradiction requires both quotes — no "vibes shifted" claims. Judge momentum on their words across time, not on call frequency or length.
</constraints>

<output_format>
The dated timeline, the contradiction pairs, then a five-sentence "state of the deal" summary a skip-level manager could read cold.
</output_format>

<context>
[PROSPECT COMPANY] sent a security and vendor questionnaire — [NUMBER OF QUESTIONS] questions, wanted back by [RETURN DATE]. The AE needs it triaged before anyone touches it manually.
</context>

<questionnaire>
[PASTE THE FULL QUESTIONNAIRE]
</questionnaire>

<knowledge_base>
[PASTE YOUR SECURITY DOCS, PAST APPROVED ANSWERS, AND CERTIFICATION SUMMARIES]
</knowledge_base>

<task>
Sort every question into three buckets: (a) answerable now from the knowledge base — draft the answer; (b) needs a subject-matter expert — name which function: security, legal, or infrastructure; (c) red flag — the honest answer may hurt the deal; explain why and suggest how the AE should address it live rather than in writing.
</task>

<constraints>
Same rule as an RFP: knowledge base or nothing — no improvised security claims, ever. Red-flag guidance coaches the delivery, never the facts. If the knowledge base contradicts itself, surface the contradiction instead of picking a side.
</constraints>

<output_format>
Three labeled lists; drafted answers inline for bucket (a); a one-line routing note for each item in bucket (b).
</output_format>

<context>
An executive sponsor meeting for [DEAL NAME] at [PROSPECT COMPANY] happens in [TIME UNTIL THE MEETING]. The exec attending: [EXEC NAME AND TITLE]. Objective: [MEETING OBJECTIVE]. The deal record is a folder, not a story — fix that.
</context>

<doc_pack>
[PASTE EVERYTHING — CALL NOTES, EMAIL THREADS, THE PROPOSAL, INTERNAL DISCUSSION EXCERPTS]
</doc_pack>

<task>
Compress the pack into a one-page brief the exec reads in three minutes: where the deal stands; what the prospect's business is trying to do, in their own words; what we are asking the exec to actually do in the room; the landmines — open issues the prospect may raise; and the one number that matters most.
</task>

<constraints>
350 words maximum. No sales jargon the exec would wince at — write it like an operator. Landmines must come from the pack; if the record shows none, write "no open issues surfaced in the record" rather than inventing caution.
</constraints>

<output_format>
Five labeled sections in the order above.
</output_format>

<context>
Procurement at [PROSPECT COMPANY] returned a redlined contract for [DEAL NAME]. The rep is not a lawyer and does not pretend to be. Deal context: [VALUE, CLOSE DATE, ANYTHING TIME-SENSITIVE].
</context>

<original_contract>
[PASTE THE ORIGINAL TERMS, OR THE SECTIONS THAT CHANGED]
</original_contract>

<redlined_version>
[PASTE THE REDLINED VERSION]
</redlined_version>

<task>
Compare the two versions and report: every substantive change in plain language; what each change means commercially for the seller; a severity rating — routine, negotiate, or escalate to legal; and which changes commonly trade against each other, so the rep walks into the negotiation holding pairs instead of a flat list.
</task>

<constraints>
Open with one line stating this is deal preparation, not legal advice. Never rate a liability, indemnity, or intellectual-property change below "escalate to legal". Quote the changed language for anything rated negotiate or higher.
</constraints>

<output_format>
A change table — clause · what changed · plain-language meaning · severity — followed by the suggested trade pairs.
</output_format>