How To Scrape UNLIMITED Business Leads Using Claude
AI automation consultant Abby demonstrates how to use Claude's Co-work agent on the Max plan to scrape hundreds of business leads across any industry and location, complete with contact details and personalized LinkedIn connection requests. The workflow involves giving Co-work a natural language prompt, refining outputs through conversation, and saving the interaction as a reusable 'skill' to avoid repeating instructions. Abby argues this approach replaces expensive third-party lead generation tools costing $50+/month.
Summary
In this tutorial, host Sabrina interviews Abby, an AI automation consultant with over a decade of fintech experience, who demonstrates how to use Claude's Co-work desktop agent to generate unlimited business leads at scale. The video opens by showcasing a pre-built spreadsheet of 10 London fintech startups, each with company name, key contact, title, address, office phone, personal work email, LinkedIn profile URL, and notable achievements used in personalized connection requests.
Abby walks through the process step-by-step: navigating to Co-work (distinct from Claude Chat and Claude Code), clicking 'New Task,' and entering a natural language prompt requesting 20 fintech prospects from New York, specifying a save folder. Co-work then autonomously browses the internet, finds leads, and returns results within 5–7 minutes for a 20-lead request. Abby notes that Co-work can ask clarifying questions — for example, distinguishing between commercial and residential real estate — to refine results.
A key refinement moment Abby describes is pushing back when Co-work initially returned generic '[email protected]' emails rather than personal work emails. After prompting it to correct this, it retrieved individual work emails and also added notable achievements for each contact, which Abby uses to personalize LinkedIn connection requests. She reports that this personalization significantly increases response rates.
Abby then explains the 'skill' feature: after a successful task interaction, she instructed Co-work to convert the entire workflow into a reusable skill. This skill captures all instructions — researching companies, finding decision makers, pulling contact details, writing personalized LinkedIn requests, and delivering results as a formatted Excel spreadsheet — so future requests can be executed without repeating every step. She notes the LinkedIn sending step is marked optional in the skill so she retains manual control over when connections are actually sent.
On plan requirements, Abby clarifies that the Pro plan hit usage limits during her testing, prompting her to upgrade to the Max plan. She also notes that extended thinking can be toggled off for this use case, and Sonnet 4.6 is sufficient to achieve the results shown. She has used this workflow across dozens of industries including real estate, dental clinics, law firms, and fintech, and has scraped hundreds of leads total. The video concludes with Abby encouraging viewers to test limits and scale up to 50 or 100 leads per run.
Key Insights
- Abby found that Co-work initially returned generic '[email protected]' emails, but after she pushed back in the conversation, it went back and retrieved personal work emails and notable achievements for each contact, demonstrating that iterative prompting within a session meaningfully improves output quality.
- Abby argues that converting a successful Co-work task interaction into a 'skill' eliminates the need to re-explain instructions for future lead generation runs — the skill encapsulates the entire workflow from company research to formatted Excel delivery.
- Abby deliberately marked the LinkedIn connection-sending step as optional in her saved skill, because she wants to retain manual control over when outreach is sent rather than allowing Co-work to send on her behalf automatically.
- Abby reports that the Pro plan hit usage limits during her testing and told her to wait until a specific time to continue, which led her to upgrade to the Max plan to avoid interruptions during lead scraping workflows.
- Abby claims that sending personalized LinkedIn connection requests referencing each prospect's notable achievements — sourced automatically by Co-work — results in higher response rates because recipients perceive that genuine research effort was made.
Topics
Full transcript available for MurmurCast members
Sign Up to Access