AI prompts for sales: examples to use in Claude, connected or not to your data

There are two types of sales prompts for Claude. The first works in any conversation, with no real data from your operation behind it. The second only works because Claude is connected to the real sales funnel, with real leads and deals to query. Both have value, but they solve different problems. This article separates ready-made examples of both types and explains what changes in the answer when there is real data behind the question.

The question is the same, the answer changes with the context

Ask Claude “which leads should I prioritize this week” with no context at all, and the answer comes as a generic script: consider the contact's job title, the size of the company, how much they have already interacted with your website. It is a good starting point, but it is still theory, because Claude is answering based on general sales best practices, not with your funnel in front of it.

Now ask the same thing to a Claude connected to the CRM and to the operation's real funnel, and the answer swaps theory for a list: the deals that have been stuck the longest at the proposal stage, the lead who opened the follow-up email more than once this week, the account that visited the pricing page yesterday. The question did not change. What changed was the context Claude received before answering, and it is that context that separates a generic sales prompt from a sales prompt that is useful day to day.

That difference explains why so many ready-made sales prompts, however well written, deliver an answer that is correct and useless at the same time: correct because it follows real sales best practices, useless because it knows nothing about your specific business. Without real data, every sales prompt becomes a well-written exercise in guesswork, and the answer is only as good as the context it receives.

Prompts that work in any Claude, with no connection at all

Even without connecting Claude to any data, there are prompts that really help. The difference is that you need to paste the context manually, prompt by prompt.

Preparing an objection before the call:
“Act as an experienced sales consultant. My prospect is a [job title] at a [size and industry] company, and their objection was [describe the objection, for example: high price compared to the competition]. Give me three different ways to respond to this objection, each with a different tone: direct, consultative, and backed by market data.”

This prompt helps you get to the conversation better prepared, but it remains generic: Claude does not know the real history of that deal, only what was described in the prompt. The more detail you include about the prospect, the better the answer, because the entire context depends on what is written there.

Summarizing and organizing the follow-up of a call:
“Here is the transcript of a sales call: [paste the transcript]. Summarize the main points, list the objections raised, and write a follow-up email that picks up the agreed next step.”

This prompt works well because the context, the transcript, is already inside the question itself. It is an example of how giving context manually solves a good part of the problem, even with no connection to real data.

Assessing whether a lead fits the ideal profile:
“Assess this lead based on this ICP: [describe the criteria, for example: a company that already invests in paid media and uses a CRM]. Here is the information I have about the contact: [paste the available data]. Does this lead fit the profile? What is still missing to confirm?”

Without a connection to real data, Claude only analyzes what was pasted into the prompt. The result depends entirely on how much context the salesperson remembered to include, and it is easy to forget a detail that changes the answer.

Building a cold prospecting sequence:
“Create a sequence of 4 cold prospecting emails for [job title] at [industry] companies. The first email should present the problem we solve, the second should bring a concrete use case, the third should respond to a common objection about [objection], and the fourth should be a polite closing email, with no pressure.”

This type of prompt saves a lot of writing time, but the result is still a skeleton: it is up to the salesperson to review the tone, confirm that the problem described really is that segment's pain, and adjust the use case to something the company has actually solved.

What changes when Claude is connected to the real funnel

There is a way to skip the step of pasting information manually with every question: connecting Claude to a real data source, through a protocol called MCP (Model Context Protocol). In practice, the agent starts seeing the real sales funnel, and each prompt becomes a query on top of the operation's own real data, not a hypothetical example.

The same lead qualification prompt, made for a connected Claude, changes shape. Instead of pasting information one piece at a time, the question becomes: “which leads from the last 7 days already fit my ICP and have not yet had any contact from the sales team?” The answer comes out with real names, companies, and behavior, without the salesperson having to paste anything beforehand.

The same goes for the objection prompt: instead of describing the prospect manually, the question can become “bring the history of this deal's interactions with us and suggest how to approach the price objection based on what they have already seen of the product”. The answer stops being generic because it starts from what really happened, not from a summarized description in the prompt.

This same logic applies to deciding which lead to prioritize first in the day's queue, a topic this article on lead prioritization explores in more depth. With the real funnel available, the question swaps “what prioritization theory says” for “who, in my queue today, already fits this criterion”.

Laiki's MCP is an example of how this connection works in practice: it unifies leads, deals, and paid media in a single infrastructure that the agent queries, without anyone having to copy a spreadsheet or export a report before asking.

How to choose the right prompt for each moment

The two types of prompt coexist well, because they serve different stages of the work. Without a connection to data, Claude helps prepare, write, and organize: objection script, call summary, email draft, prospecting sequence. With the real funnel connected, Claude helps diagnose and prioritize: which deals are stuck, which leads have gone cold, which account deserves attention today.

The Laiki Prompt Guide brings together both types of example, organized by area (Marketing, Sales, and Behavior) and by data source (GA4, Meta Ads, Google Ads, LinkedIn Ads, Segments, and Leads).

Download the Laiki Prompt Guide: 70 ready-made prompts, including the Sales ones, organized by area and by data source. Access the complete guide.

If you want to see the complete list of the 70 prompts, including the marketing and behavior examples besides the sales ones, the article AI prompts for marketing and sales: 70 ready-to-use examples details each category.

Frequently asked questions

Do these prompts work only in Claude?
Most work in any conversational AI agent, including ChatGPT and Gemini. The real difference is not in the brand of the agent: it is in whether or not it has a connection to the real data of your operation via MCP or another integration.

Do I need to know how to code to connect Claude to my data?
No. The connection via MCP works like installing a connector, without requiring code. The step-by-step for installing Laiki's MCP in Claude and ChatGPT is detailed in this installation tutorial.

Do I need to already be a Laiki customer to use these prompts?
No. The prompts without a connection work in any Claude, with any operation. The prompts that query real data require a connected data source, and Laiki's MCP is one of the ways to make that connection, but the reasoning applies to any agent connected to your funnel.

Does this replace the salesperson?
It does not. The agent speeds up the stage of gathering and organizing information, which used to take a good part of the day. The conversation, the negotiation, and the decision remain the salesperson's, only with less time wasted manually building context.

Start with the right prompt for today

If what is missing right now is organizing a call, preparing an objection, or building a prospecting sequence, the prompts without a connection already help, without requiring any setup. If the goal is to stop explaining the funnel from scratch with every question, it is worth connecting Claude to the operation's real data and using the Sales prompts from the complete guide.

Download the Laiki Prompt Guide: Access the complete guide and start with the prompt that already solves the decision you need to make today.