AI outbound engine: Clay enrichment + LLM first lines
Rebuilding my manual Apollo → Instantly cold email workflow AI-first: Clay enrichment, ICP scoring and Claude-written openers, pushed into Instantly.
The problem
At Maildoso I ran cold email campaigns to lists of up to 100,000 prospects. List building, enrichment and personalization were the slowest parts, and generic first lines hurt reply rates.
The plan
- Pull an ICP list (title, company size, industry) from Apollo into Clay.
- Waterfall enrichment: company website, recent news, tech stack, hiring signals.
- Score every lead against the ICP and drop the bad fits before sending.
- Prompt Claude to write a one-line, fact-based opener per lead, with guardrails so it can’t make things up.
- Push qualified leads and their openers into an Instantly campaign.
What I’ll measure
- Hours saved per 1,000 leads compared with the manual process
- Reply rate against a non-personalized control group
- Cost per enriched lead
Results
Coming soon.
Next build →
Webinar → content pipeline in n8n