Research companies and fill a CRM for outreach
How to Build Your First AI Sales Engine With Claude Code. Task: Jonas Braadbaart built a sales engine using Claude Code to automate repetitive sales admin tasks. The workflow pulls company data from Apollo, filters for Ideal Customer Profile (ICP) fit, conducts deep research via Perplexity, creates records in Attio CRM, and drafts outreach messages. The human manually prescreens the initial list, reviews the generated reports, and edits all outreach drafts before sending, ensuring the agent handles heavy lifting while the human retains final decision-making authority.; tools: Claude Code. Human role: The original post describes the human role this way: The Agent Drafts Outreach And Stops Before Sending; limitations: The workflow relies on specific integrations: Apollo for data, Attio for CRM (due to its mature MCP server), and Perplexity for research. Fully automated cold outreach was previously found to be ineffective and robotic, necessitating a strict human-in-the-loop review step. Costs and results are specific to Jonas's solo operator context and may vary with different tool stacks or list sizes..
Tools
Published on Agent Work Atlas:
Overview
How to Build Your First AI Sales Engine With Claude Code
Workflow
Jonas Braadbaart built a sales engine using Claude Code to automate repetitive sales admin tasks. The workflow pulls company data from Apollo, filters for Ideal Customer Profile (ICP) fit, conducts deep research via Perplexity, creates records in Attio CRM, and drafts outreach messages. The human manually prescreens the initial list, reviews the generated reports, and edits all outreach drafts before sending, ensuring the agent handles heavy lifting while the human retains final decision-making authority.
Reported result
In the first run, Jonas contacted 80 companies and secured one sale. The system reduced the time required for review and message preparation to approximately one minute per lead, compared to an estimated one to two hours manually. The total operational cost for the tools used was approximately 120 euros, excluding the existing Claude subscription.
Human role
The original post describes the human role this way: The Agent Drafts Outreach And Stops Before Sending
Limitations
The workflow relies on specific integrations: Apollo for data, Attio for CRM (due to its mature MCP server), and Perplexity for research. Fully automated cold outreach was previously found to be ineffective and robotic, necessitating a strict human-in-the-loop review step. Costs and results are specific to Jonas's solo operator context and may vary with different tool stacks or list sizes.
Try it
Act as a sales operations agent. I will provide a CSV of 20 companies from [Apollo/CRM Export] and my ICP criteria: [Insert ICP, e.g., Mid-sized service companies]. 1. Score each company against the ICP and select the top 5. 2. For each selected company, use Perplexity to research their business model, AI strategy, and key decision-makers. 3. Create a structured report for each company summarizing the research and potential collaboration angles. 4. Draft one LinkedIn message and one email for each company, referencing specific pain points found in the research. 5. STOP. Do not send or create CRM records yet. Present the reports and drafts to me for review. Wait for my feedback before proceeding to any CRM updates or sending actions. Ensure all drafts sound human and personalized, avoiding generic AI phrasing.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.