Updated September 28, 2026 | 10 min read
Updated September 28, 2026 | 10 min read
Ready to launch?
Convert strategy into pipeline
Launch your first outbound campaign with verified leads, AI sequences, and multichannel outreach from one platform.
18 Claude Prompts for Sales Teams (Copy-Paste Templates for 2026)
<article>
Last week I watched a rep paste “write a cold email to a VP of Sales at a fintech company” into Claude, read the output, and send it. It opened with “I hope this email finds you well.” It closed with an offer to “explore synergies.” He got nothing back, and he blamed the model. The prompts are the problem. Claude will happily write forgettable copy if forgettable copy is all you asked for.
Last week I watched a rep paste “write a cold email to a VP of Sales at a fintech company” into Claude, read the output, and send it. It opened with “I hope this email finds you well.” It closed with an offer to “explore synergies.” He got nothing back, and he blamed the model. The prompts are the problem. Claude will happily write forgettable copy if forgettable copy is all you asked for.
In this post, I’ll give you 18 copy-paste prompts across six categories, plus the prompt structure I use to get output that’s actually sendable. You’ll get cold emails and research briefs, plus objection responses you can use without rewriting them line by line. At the end, I’ll show you how to run these against live prospect data instead of copy-pasting between tabs.
What makes a Claude sales prompt actually work
To get non-generic sales output from Claude, you assign a specific expert role, feed it precise context (a target company, a trigger event), and strictly ban AI filler phrases. That’s the formula: role + context + constraints.
When you type “write me a cold email,” Claude has nothing to work with. So it defaults to safe, forgettable copy that sounds like every other AI-generated message in your prospect’s inbox. The fix is giving Claude the same briefing you’d give a new SDR on their first day: who the prospect is and what just happened at their company, plus a clear picture of what good output looks like.
Use XML tags to structure context
XML tags like
<context>, <company>, and <trigger_event> help Claude parse information cleanly. Instead of dumping everything into one paragraph, you separate the pieces so Claude knows what’s what. This tagging pattern comes straight from Anthropic’s prompting documentation, which covers general and model-specific prompting guidance.<company>Acme Corp, Series B fintech, 200 employees</company>
<trigger_event>Just hired VP of Sales last week</trigger_event>
<your_product>Outbound automation platform</your_product>
Without this structure, Claude might confuse your product with the prospect’s company, or miss the trigger event entirely. With it, every piece of information has a clear label.
Give Claude a role and a clear goal
Instructions like “Act as a senior B2B sales rep targeting mid-market SaaS” change output quality dramatically. An enterprise AE writes differently than an SDR. A sales coach frames objection responses differently than a copywriter. Claude mirrors whichever persona you specify, so be specific about who you want it to be.
Show examples of winning output
Claude learns from examples you provide. Paste a cold email that actually got replies and Claude will mirror the style, length, and tone. This is called few-shot prompting, and it’s the fastest way to get output that sounds like your team rather than a chatbot.
Even one example helps. Two or three help more. The point is showing Claude what “good” looks like for your specific use case.
Set constraints on length and tone
Without constraints, Claude defaults to verbose, safe copy. With them, you get something you can actually send. Four constraints do most of the work:
- Cap the word count. “Under 75 words” forces Claude to cut the fluff.
- Ban the obvious tells. “No ‘I hope this finds you well’” kills the phrase that flags your email as AI-written before anyone reads sentence two.
- Direct the tone. “Sound like a peer, not a vendor” shapes the whole voice.
- Control the formatting. “No markdown, no bullet points” stops Claude from returning something that looks like a slide deck when you asked for an email.
Best Claude prompts for prospecting and account research
Account research is where Claude saves the most time. These three prompts replace most of the tab-switching you do before a call.
Deep account research brief
Act as a B2B researcher. Based on the following inputs, create a research brief:
<company_website>[URL]</company_website>
<trigger_event>[Recent news or signal]</trigger_event>
<your_value_prop>[One sentence]</your_value_prop>
Output:
- Company summary (2 sentences max)
- 3 likely pain points based on their industry and trigger
- 2 talking points for a cold call
- One cold email opening line
Constraints: Under 150 words total. No filler phrases.
This turns a research task you’d normally spread across half a dozen browser tabs into a single pass. It works best when the inputs are already enriched, which is why I run it on top of records pulled from lemlist’s 600M+ lead database rather than on a name and a guess.
Buying committee mapping
Act as a sales strategist. Based on this company profile:
<company>[Name, size, industry]</company>
Identify the likely buying committee for [your product category].
For each role, list:
- Their primary concern
- How they evaluate vendors
- One question they'll ask
Output as a simple table.
This prompt pairs well with intent data. When you know a company just posted a VP of Sales role, you can map the committee before they even fill the position. If you’re tracking signals like hiring changes or funding rounds, lemlist’s Intent Signals can surface the accounts worth researching in the first place.
Tech stack and hiring signal scan
Act as a GTM researcher. Score this account on fit and timing:
<company>[Name, size, industry, HQ]</company>
<tech_stack>[Tools detected]</tech_stack>
<hiring_signals>[Open roles posted in last 60 days]</hiring_signals>
<our_icp>[Two sentences describing our ideal customer]</our_icp>
Output:
- Fit score out of 10 with one line of reasoning
- What their tech stack tells you about how they buy
- What their open roles reveal about current priorities
- One outreach angle that ties the stack and the roles together
- Verdict: pursue now, nurture, or skip
Constraints: Under 120 words. No hedging language.
Tech stack and hiring data are filters in lemlist’s Lead Database, so you can build the list first and score it second. Feed Claude twenty accounts in one go and you’ll have a prioritized call list before your coffee goes cold.
Best Claude prompts for cold email outreach
Cold email is the highest-demand prompt category. Here’s what actually works.
First touch cold email
Act as a senior B2B copywriter. Draft a cold email with exactly 4 sentences:
1. Reference a peer challenge in their industry
2. Introduce a solution (not your product name)
3. Include one specific metric or outcome
4. End with a low-friction CTA
<prospect_company>[Name]</prospect_company>
<trigger_event>[What just happened]</trigger_event>
<your_solution>[What you help with]</your_solution>
Constraints: Under 75 words. Sound human, not corporate. No "I hope this finds you well."
The structure here matters. Sentence one shows you understand their world. Sentence two introduces a solution without pitching. Sentence three adds credibility with a specific number. Sentence four makes it easy to say yes.
Signal-based cold email
Act as an SDR who just saw this intent signal:
<signal>[Funding round / job posting / tech adoption]</signal>
<prospect>[Company and role]</prospect>
Write a cold email that references the signal naturally in the first line,
connects it to a relevant pain point, and asks for 15 minutes.
Constraints: Under 70 words. No generic congratulations.
Intent signals turn cold outreach into warm outreach. The email feels timely because it is timely. A funding round means they have budget. A VP of Sales hire means they’re scaling. A new tech adoption means they’re solving a problem you might also solve.
Five-step cold email sequence
Act as a B2B outbound strategist. Build a 5-email cold sequence for:
<icp>[Role, company type, main pain]</icp>
<our_offer>[One sentence]</our_offer>
<proof>[One customer result with a number]</proof>
For each email, output:
- Day in the sequence
- Subject line (under 5 words, lowercase)
- Body (under 80 words)
- The single angle it uses (pain, proof, insight, peer story, or breakup)
Rules: Every email must work if the prospect never read the previous one.
No "just following up." No repeated opening lines across the sequence.
One prompt gets you the whole sequence, which you then load into lemlist as an automated sequence with delays and a stop-on-reply rule. The angle-per-email constraint is the part people skip. Without it, Claude writes the same email five times with different verbs.
Best Claude prompts for LinkedIn prospecting
LinkedIn runs on different constraints than email: shorter messages, more casual tone, hard character limits. These prompts account for that, and they follow the order that makes LinkedIn prospecting feel natural rather than transactional. Comment on their post, then send the connection request, then open the DM.
Warm comment before connecting
Act as a peer in [industry] reading this LinkedIn post:
<post>[Paste the post]</post>
<author>[Name, role, company]</author>
Write a comment that adds one specific idea or counterpoint the author
didn't cover. Reference something concrete from the post.
Constraints: Max 2 sentences. No compliments. No emojis. No exclamation marks.
Never mention my company or what I sell.
The comment is the warm-up. By the time your connection request lands, your name is already familiar. lemlist can automate the feed engagement step so the comment goes out before the request in the same sequence.
Connection request note
Write a LinkedIn connection request (under 300 characters) for:
<prospect>[Name, role, company]</prospect>
<shared_context>[Mutual connection, content they posted, or industry overlap]</shared_context>
Do not pitch. Reference the shared context only.
The only goal here is getting accepted. Leave out the ask and the value prop, and give them a reason to connect that reads as genuine.
LinkedIn DM after connection
Act as a sales rep who just connected with this prospect:
<prospect>[Name, role]</prospect>
<connection_reason>[Why they accepted]</connection_reason>
Write a first DM that bridges from the connection reason to a soft value prop.
No hard pitch. Under 100 words.
Best Claude prompts for discovery calls
Discovery prompts earn their keep before the call and right after it.
Role-specific discovery questions
Act as a MEDDIC-trained enterprise AE. Generate discovery questions for:
<prospect_role>[Title and function]</prospect_role>
<company_context>[Size, industry, trigger]</company_context>
<our_category>[What we sell]</our_category>
Output 8 questions grouped as:
- Current state (what they do today)
- Pain and cost (what it's costing them)
- Decision process (who signs, what blocks)
- Vision (what good looks like in 6 months)
Rules: No yes/no questions. No questions I could answer myself with a Google search.
That last constraint is the one that changes the output. Ban questions you could research yourself and Claude stops handing you “so, how big is your sales team?”
Pre-call research brief
Act as a B2B researcher preparing an AE for a discovery call.
<company>[Name, website]</company>
<prospect_role>[Title]</prospect_role>
<trigger>[Why they took the meeting]</trigger>
Output:
- Company summary (2 sentences)
- 3 likely pain points for this role
- 2 discovery questions specific to their situation
- One insight to share that builds credibility
Under 150 words.
Post-call CRM summary
Act as a sales ops analyst. Based on these raw call notes:
<notes>[Paste notes]</notes>
Output a structured CRM update:
- Key pain points (bullet list)
- Stakeholders mentioned
- Next steps agreed
- Recommended deal stage
- Red flags (if any)
This prompt saves 10 minutes after every call and keeps your CRM actually useful. Paste your messy notes, get a clean summary.
Best Claude prompts for objection handling
Good objection responses acknowledge the concern, then pivot. Here’s how to generate them.
Price objection
Act as an enterprise sales coach. Create two rebuttals (under 30 words each)
for this objection: "It's too expensive for us right now."
Method 1 - Empathy & Pivot: Acknowledge the concern, then redirect to ROI.
Method 2 - Cost of Inaction: Quantify what the status quo costs them.
Two methods give you options. Empathy & Pivot works when the prospect needs to feel heard. Cost of Inaction works when they haven’t thought through what doing nothing costs them.
Competitor objection
Act as a sales coach. The prospect said: "We already use [competitor name]."
Create two responses:
1. A question that uncovers what's not working
2. A statement about switching cost vs. opportunity cost
Under 40 words each. No trash-talking the competitor.
“Not the right time” objection
Act as a sales coach. The prospect said: "This isn't the right time for us."
First, list the 3 things this objection usually hides (no budget, no priority, no authority).
Then write one response for each, under 35 words.
Each response must end with a question that surfaces which one is true.
Constraints: No pressure tactics. No fake urgency. Accept a "no" gracefully if that's the answer.
I’ll concede the point most rebuttal frameworks avoid: sometimes it genuinely isn’t the right time, and the smart move is a calendar reminder for Q2 rather than a fourth rebuttal. Claude is good at telling the difference if you let it, which is exactly what that last constraint is for.
Best Claude prompts for follow-up and re-engagement
Most pipeline dies in the gap between “sounds interesting” and silence. These prompts are for that gap.
Follow-up cadence after no reply
Act as an outbound strategist. Design a 4-touch follow-up cadence for a prospect
who opened my first email twice but never replied.
<original_email>[Paste it]</original_email>
<prospect>[Role, company, trigger]</prospect>
For each touch, output: channel (email or LinkedIn), day, angle, and the message.
Every touch must add new information. Ban "just checking in" and "bumping this up."
Under 60 words per message.
Mix the channels here. A follow-up that arrives on LinkedIn after two emails reads as persistence; a fourth email reads as automation.
Re-engagement email for deals gone dark
Act as an AE reopening a stalled deal.
<last_contact>[Date and what was discussed]</last_contact>
<why_it_stalled>[Known or suspected reason]</why_it_stalled>
<what_changed>[New feature, new case study, new trigger at their company]</what_changed>
Write a re-engagement email that leads with what changed, not with an apology
for the gap. Give them an easy way to say "still not now."
Constraints: Under 80 words. No guilt. No "circling back."
The trigger is what makes this land. A new funding round, a new head of department, a competitor they just lost a deal to. When lemlist flags one of those signals on a closed-lost account, this prompt writes the email before the signal goes stale.
Breakup email
Act as a senior AE writing a final email before closing the file.
<prospect>[Name, role, company]</prospect>
<history>[Number of touches, what was discussed]</history>
Write a breakup email that states you're closing the loop, leaves one piece of
value behind, and makes replying optional.
Constraints: Under 50 words. No passive aggression. No "I'll assume you're not interested."
Claude vs ChatGPT for sales prompts
Both work. The differences are real, and smaller than the internet suggests.
Factor | Claude | ChatGPT |
|---|---|---|
Context window | 1M tokens on current Opus and Sonnet models per Anthropic’s model docs, enough for full call transcripts | Smaller, improving |
Writing style | More natural prose out of the box | Can feel templated |
lemlist MCP support | Supported | Supported |
No-setup option | Claude Skills runs lemlist workflows in the browser with zero config | No browser-native equivalent |
Best for | Long transcripts, sequence writing, anything you’ll send as-is | Quick one-off tasks and fast iteration |
lemlist’s own help center now documents how to run the MCP from Claude or ChatGPT, so the integration question is mostly settled. The remaining Claude-specific advantage is Claude Skills. Most sales teams I talk to end up using both models anyway and switching based on the task (and occasionally based on which tab is already open, which is a less noble reason but an honest one).
How to run Claude sales prompts on autopilot
Manual prompting works for testing. It doesn’t scale past a few dozen prospects.
The Model Context Protocol (MCP) is an open standard that lets Claude connect to live sales data and execute actions instead of only generating text. With lemlist MCP, a single prompt can search 600M+ leads, enrich contacts with verified emails, generate personalized outreach, and launch campaigns.
Connect via OAuth in under 2 minutes by adding lemlist as a custom connector, or use an API key if your security setup requires it. If you’d rather skip the client setup entirely, Claude Skills is the no-setup path: type your request in the browser and Claude runs it directly, with the same power and zero config.
What used to take 45 minutes of tab-switching becomes one prompt: “Find 50 SaaS CMOs who raised funding in the last 30 days, enrich their emails, and draft personalized cold emails referencing the funding.”
- Query 600M+ contacts using natural language filters instead of building segments by hand.
- Pull verified emails and phone numbers automatically as part of the same request.
- Generate personalized messages that use real prospect fields rather than placeholders.
- Launch multichannel sequences straight from the conversation.
If writing prompts isn’t how you want to spend your afternoon, lemAgent builds the campaign from a plain conversation instead. Same outcome, no prompt engineering.
Frequently asked questions
Is Claude better than ChatGPT for sales prompts?
For anything you’ll send without editing, Claude’s default prose tends to need less cleanup. For long inputs like full call transcripts, its 1M-token context window on current Opus and Sonnet models gives it room ChatGPT doesn’t always have. For fast one-off tasks, the difference barely shows. Both connect to lemlist through the MCP, so the integration is no longer a deciding factor.
Can Claude access my CRM through MCP?
Claude can reach any tool that exposes an MCP server and that you’ve connected. lemlist’s MCP covers lead search, enrichment, message generation, and campaign launch. For your CRM specifically, check whether your vendor ships an MCP server or connector. If it does, Claude can read and write to it in the same conversation it uses to draft your outreach.
Will prospects know my emails were written by Claude?
They will if you skip the constraints. The tells are the ones everyone recognizes: the “I hope this finds you well” opener, the three-paragraph structure, the word “synergies.” Ban them explicitly in the prompt, cap the word count, and feed in a real trigger event. What gets flagged as AI writing is almost always generic writing.
Do I need to set anything up to use these prompts?
No. Every prompt in this article works by pasting it into Claude. Setup only matters when you want Claude to act on live data rather than on what you paste, which is where MCP or Claude Skills comes in.
Over to you
A Claude prompt that works gives the model a specific role to play, real context about the account in front of you, and hard limits on length and tone. Start with one prompt from this article, run it against a prospect you’re actually working, and tighten the constraints until the output is sendable without edits.
If you want to stop copy-pasting between tabs and run these prompts against live prospecting data, start a 14-day free trial and connect Claude to your outbound motion directly. No credit card required.
</article>
</article>
