$ man research-prompt

Research Prompt

An AI prompt (usually in Clay) that takes enriched data (LinkedIn profile, company domain, job title) and generates personalization variables (icebreaker, pain point, service fit) for outbound campaigns.

You can't manually research 500 contacts. The research prompt does it for you. It reads the LinkedIn "About" section, scans recent posts, checks the company domain, and outputs 2-3 personalized hooks per contact. It's the automation layer between "raw data" and "usable copy."

I write one research prompt per campaign. The prompt embeds the full email bodies so the AI knows exactly where each variable appears and writes them to read naturally in context. I define strict rules for each variable (icebreaker = recent post or profile headline, no fabrication; pain_point = company-level signal, not contact-level guess). The prompt outputs JSON with fields that map to merge fields in the email. I test it on 5 rows, check for hallucinations, then scale. Research prompts are the bridge between enrichment and sending.