Most sales reps spend 2+ hours a day on research, CRM updates, and follow-up emails. Claude can compress that into minutes, but only if you set it up correctly.

I’ll walk you through the exact setup process, the workflows that actually save time, and the prompts you can copy today to start using Claude for pre-call prep, outreach personalization, and pipeline management.

You can use Claude for sales by connecting it to your CRM, email, and calendar, then building automated workflows for pre-call research, cold outreach drafting, and post-call follow-ups. The whole setup takes about 15 minutes. Once it’s running, Claude handles the repetitive research and writing that eats into your selling time.

Here’s what that looks like day to day:

The common thread here is that Claude handles the prep work so you can focus on the actual conversation.

Before we get into setup, it helps to understand which Claude product fits your workflow. Anthropic offers several options, and they serve different use cases.

Claude Chat is the conversational interface most people start with. You type a question, Claude answers. Simple.

Projects take it further. A Project is a dedicated workspace where you store ICP documents, playbooks, and tone guides. Claude remembers this context across conversations, so you don’t have to re-explain your business every time you open a new chat.

Skills are reusable instruction sets. You teach Claude how to run a specific task once (say, generating a pre-call brief), and then you can trigger that skill on demand without rewriting the prompt. Think of it as encoding your sales process into a repeatable workflow.

This is the terminal-based version for technical users. If you want to run automations on local files, process CRM exports in bulk, or build custom workflows with scripts, Claude Code is the tool. Most sales reps won’t touch this, but RevOps teams sometimes find it useful.

MCP stands for Model Context Protocol. It’s the way Claude connects to external tools like CRMs, outbound platforms, and databases.

With MCP, Claude can take action (search leads, launch campaigns, update records) rather than just generate text. lemlist MCP is one example: it exposes 40+ outbound actions to Claude, so you can run your entire prospecting workflow from a single prompt.

Claude is only as useful as the context you give it. Skip this step and you’ll get generic outputs that sound like every other AI-generated email. Garbage in, garbage out.

Connect Claude to the tools where your data lives. That means your CRM (HubSpot, Salesforce), email (Gmail, Outlook), calendar, and call recorder (Gong, Fathom, Fireflies). Connections happen in Settings > Connectors or through MCP servers.

Claude needs to know who you sell to. Upload a document with your ideal customer profile: firmographics, common pain points, typical objections, and decision-making criteria. The more specific you are, the better the output.

Include tone guidelines, email length constraints, phrases to avoid, and CTA preferences. Claude will mirror whatever style you define. If you want two-sentence openers with no exclamation marks, say so explicitly.

Tell Claude what it can’t do: make up pricing, promise features that don’t exist, skip compliance language. Guardrails protect you from outputs that could create problems downstream.

Here’s the actual setup process. Most teams can complete this in a single sitting.

Download the Claude Desktop app from claude.ai. Once installed, create a new Project and name it something like “My Deals” or “Outbound.”

Then add your ICP document, sales playbook, and tone guide to the Project knowledge. Claude will reference these files in every conversation within that Project.

Go to Settings > Connectors and link your CRM, email, and calendar. If you want Claude to execute actions (not just read data), you’ll connect via MCP.

For example, lemlist MCP lets Claude search leads, enrich contacts, and launch campaigns directly. The connection requires an API key and a simple JSON config file.

Skills are point-and-click reusable instructions you create in the Claude interface. CLAUDE.md is a markdown file for Claude Code users that encodes the same information.

Either approach works. The goal is to define your sales process once so Claude runs it consistently every time.

In your Project settings, add a system prompt that defines Claude’s role, output format preferences, tone constraints, and escalation rules. Keep it under 500 words.

Example: “You are a sales research assistant. Always cite your sources. Keep emails under 100 words. Never mention competitor pricing.”

Before going live, run one pre-call research brief or one follow-up email. Compare the output to your quality bar. If something’s off, adjust your instructions and test again. This iteration step is where most of the value comes from.

Different roles have different daily tasks. Here are the workflows that save the most time, organized by what you’re actually trying to accomplish.

Paste a LinkedIn URL or company name and ask Claude for a one-page brief. You’ll get recent news, potential pain points, and suggested talking points. Target: under 60 seconds to generate, under 60 seconds to read.

Feed Claude a list of target accounts. It researches trigger events (funding, hiring, product launches) and drafts sequences tailored to each account.

This workflow becomes more powerful when connected to an outbound platform via MCP. Claude can then execute the send, not just draft the message.

Paste raw call notes or a transcript. Claude generates a follow-up email plus structured CRM notes: next steps, deal stage, key decision-makers, concerns raised. You review, edit, and send.

Give Claude your CRM export or connect it directly. It generates a report with deal movement, at-risk deals, and recommended actions. Sales managers use this to prep for pipeline meetings in minutes instead of hours.

Claude identifies deals with no recent activity and generates re-engagement messages based on the last conversation context. Useful for reviving deals that slipped through the cracks.

Here are ready-to-use prompts. Copy, paste, and adjust the placeholders.

Executing this sequence requires an outbound platform. Claude drafts; a tool like lemlist sends.

Prompts alone generate text. MCP lets Claude take action.

With an MCP connection, you can tell Claude: “Find 50 SaaS founders in fintech, enrich with verified emails, and add to my Q3 outreach campaign.” Claude searches the database, enriches the contacts, and adds them to the campaign. No tab-switching, no copy-paste.

Setup is straightforward. Generate an API key from your outbound platform, add the MCP server URL and key to your Claude config (a simple JSON file), and you’re connected.

lemlist MCP exposes 40+ actions, from lead search across 450M+ contacts to campaign launch and optimization. The entire workflow (research, personalize, send, optimize) runs from a single prompt.

A few patterns kill output quality. Here’s what to watch for:

Setting realistic expectations saves frustration later.

Claude can’t make calls. It can prep and script, but a human or dialer handles execution. It can’t send emails autonomously unless connected via MCP to a tool that sends. It can’t access real-time data without connections.

And it can’t replace relationship-building. The human parts of sales remain human. Claude handles the prep work so you can focus on the conversation itself.

Claude generates the strategy and content. Executing multichannel outreach at scale requires an outbound platform.

lemlist MCP bridges that gap. Claude searches lemlist’s 450M+ contact database, enriches leads with verified emails and phone numbers, and launches campaigns across email, LinkedIn, and calls. Intent signals from lemlist can feed Claude prompts for signal-based personalization.

The entire workflow (research, personalize, send, optimize) runs from a single prompt.

Claude has native Projects for persistent context, Skills for reusable workflows, and MCP for tool connections. ChatGPT has similar features (GPTs, plugins), but the implementation differs. Your choice often depends on which integrations you need.

Claude handles research, drafting, and data formatting well. It reduces admin time so reps can spend more time in conversations. Output quality depends on the context and instructions you provide.

Claude can handle parts of the SDR workflow (research, personalization, follow-ups) but can’t make calls, build relationships, or handle live objections. Think of it as SDR augmentation, not replacement.