Most pages ranking for Claude prompts for sales hand you a list of prompt text and stop. The prompts are usually fine. What is missing is the part that decides whether they help you close anything: where the prompt lives, what context it can see, and whether you rebuild that context from scratch every time you open a new chat. Claude's useful traits for a seller are a long context window, steady handling of pasted documents, and Projects that hold reference material across conversations. A flat prompt list ignores all three. This page gives you prompts by deal stage, then shows how to run them so the same account background is not pasted five times a week.
Why a Claude prompt for sales is not a ChatGPT prompt with a new name
Three differences matter in practice. First, length. You can paste an entire discovery transcript, a 40-page RFP and last quarter's win-loss notes into one conversation and ask questions across all of it. Prompts written for a short window tend to ask you to summarise first, which throws away the detail you needed. Second, document behaviour. Claude tends to stay anchored to what you gave it and will say when the source does not answer the question, which is the behaviour you want when the output is going to a customer. Third, and most useful, Projects. A Project holds files and standing instructions that every conversation inside it can see, so account context becomes something you set up once rather than paste repeatedly.
The practical consequence: a good Claude prompt for sales is usually shorter than the ChatGPT equivalent, because the context that a ChatGPT prompt has to carry inline lives in the Project instead.
Prospecting and account research prompts
Research prompts fail when they ask for a company summary. You do not need a summary. You need a reason to reach out this week.
Account brief: Here is the company's last two earnings calls and their careers page. Identify the three operating priorities leadership repeated across both calls, the evidence for each, and one hiring pattern that suggests where budget is moving. Flag anything you are inferring rather than reading directly.
Trigger hunt: Given these five accounts and what we sell, rank them by how likely they are to be in-market in the next 60 days. For each, name the specific signal you are using and the one that would change your ranking if it were absent.
More of these sit in Account Research and Buyer Intelligence.
Discovery call prep prompts
Pre-call brief: You are preparing me for a first call with this persona at this company. Using the attached research, write a one-page brief: the pain hypothesis, the three questions that would confirm or kill it, the two things I should not say, and the most likely reason this call goes nowhere.
Question design: Draft eight discovery questions that would surface how this buyer measures the problem. No question may be answerable with yes or no, and none may reference our product.
The ChatGPT counterpart to this section is ChatGPT prompts for discovery calls; the prompts themselves live in Meeting Prep and Discovery.
Objection handling and negotiation prompts
Objection pressure test: Here is the objection, verbatim, and the context around it. Give me three readings of what the buyer might actually mean, ranked by likelihood, and one question that would tell me which reading is right before I respond.
Concession planning: Here is our pricing, the deal shape and what the buyer has asked for. Build a concession matrix: what we can trade, what we get back for each trade, and the point at which walking away is the better outcome. Argue the buyer's side of each trade afterwards.
See also ChatGPT prompts for objection handling.
Follow-up, multithreading and deal-advance prompts
Recap that advances: Here are my raw notes from the call. Write the follow-up email: confirm what they said in their words, name the one thing we disagreed on rather than smoothing it over, and propose a next step that is specific enough to say no to.
Multithreading: Based on this call transcript, list every person mentioned, what we know about their stake in the decision, and who is missing from the conversation entirely. For the biggest gap, draft the request I would make to my champion to get introduced.
Message-level prompts are catalogued in Outreach and Messaging.
Pipeline and forecast prompts for managers
Deal inspection: Here is the CRM export for my team's committed deals. For each, list what is documented, what is assumed, and the single missing piece of evidence that would most change my confidence. Rank the deals by how much the forecast depends on an assumption.
Forecast narrative: Using this pipeline data, write the forecast commentary I would give my VP: the number, the two deals it rests on, the risk I am carrying, and what would have to happen this week for the number to hold.
Manager prompts sit in Sales Operations, CRM and Productivity.
Stop re-pasting context: running an account in a Claude Project
This is the step the competing pages skip. A Claude Project is a container with its own files and standing instructions. Every conversation you start inside it inherits both. For a named account, that means one Project holding the discovery transcript, the mutual action plan, the security questionnaire and a short instruction block describing your product, your pricing and how you want output written. From then on, the prompts above get shorter, because the context is already there.
A workable setup for an enterprise seller: one Project per strategic account, plus one general Project holding your ICP, your objection library and your writing style. The general Project handles prospecting and message drafting. The account Projects handle everything after the first call. The instruction block is where you put the rules you are tired of repeating, such as never inventing customer names, always flagging inference, and writing at under 120 words unless asked.
When a repeated prompt is worth turning into a Claude Skill
A Skill is worth building when three things are true at once: you run the same prompt at least weekly, the output has a fixed shape you keep correcting toward, and someone else on the team should be running it the same way. Call recap formatting, MEDDPICC gap analysis and renewal risk scoring all qualify. One-off research does not. The test is not complexity, it is repetition with a standard, and the honest answer for most reps is that two or three prompts qualify, not twenty.
For the wider picture of what Claude does well in a sales week, see Claude for B2B sales workflows.
Where to keep the prompts you reuse
Projects hold context. They are not where a prompt library lives, because a prompt you can only find inside one Project is a prompt you will rewrite badly at 8am before a call. Keep the prompts somewhere you can search by stage, role and task, and paste the one you need into whichever model your company allows this quarter. That portability matters more than it sounds: the model changes, the deal stages do not. Start with a free prompt, or Browse the library →.
Frequently asked questions
Are Claude prompts for sales different from ChatGPT prompts?
The sales thinking is the same and the phrasing usually differs. Claude handles long pasted documents well, so prompts written for it can point at a transcript or an RFP rather than carrying a summary inline. Claude Projects also hold account context across conversations, which lets the prompt itself be shorter. A ChatGPT prompt will still work in Claude; it will just be doing work the Project could do for you.
What is the best Claude prompt for cold outreach?
The one that gives Claude evidence instead of adjectives. Paste the account research and ask for an opening line that references one specific, verifiable thing about the company, then ask for the email in under 90 words with a single ask. Then ask Claude what a busy buyer would delete. The second pass is where the quality comes from.
Can I save my sales prompts inside Claude?
Partly. Claude Projects hold files and standing instructions, so you can keep frequently used prompts in a Project file, and a repeated prompt with a fixed output shape can become a Claude Skill. Neither is a searchable library across stages and roles, which is why most reps keep the prompts outside the model and paste them in.
Is Claude or ChatGPT better for sales work?
Claude is generally stronger on long documents, transcript analysis and staying anchored to source material, which covers discovery synthesis, RFPs and deal review. ChatGPT has the wider tool and integration ecosystem. Most sellers do not get to choose, because the company chooses. Write your prompts so they survive that decision.
Do Claude Projects help with account research?
Yes, and mostly by removing repetition. Put the research, the call notes and the account plan in one Project, and every later question about that account starts with the model already holding the history. The gain is not a better first answer, it is not paying the setup cost again on the twelfth question.