AI-powered Outbound Cold Email Personalization at Scale
Cold outreach doesn't scale past 20 emails a day... until it does.
I built an AI-powered outbound personalization system for a mock client, Verve Outbound — a lean 2-person B2B SaaS team who needed to send cold emails that don't read like a mail merge, without hiring an SDR.
What it does:
→ Reads a prospect list (name, company, website)
→ Scrapes each company's site for one real, verifiable detail
→ AI writes a personalized opening line from that detail — not a generic hook
→ Auto follow-ups that stop the moment someone replies (thread-ID matching, not just a timer)
One decision I'm glad I made early: I skipped LinkedIn scraping entirely. It violates their ToS and risks account bans — not worth it for marginal personalization gains when a company website gives you enough signal anyway.
Reply detection is built on Gmail thread-ID matching rather than reading email content — cleaner, more reliable, and it's the same mechanism you'd use in production. (Tested manually for this demo; the logic is identical either way.)
This isn't about replacing an SDR — it's about removing the manual bottleneck so a 2-person team can actually run outbound at all.
I built an AI-powered outbound personalization system for a mock client, Verve Outbound — a lean 2-person B2B SaaS team who needed to send cold emails that don't read like a mail merge, without hiring an SDR.
What it does:
→ Reads a prospect list (name, company, website)
→ Scrapes each company's site for one real, verifiable detail
→ AI writes a personalized opening line from that detail — not a generic hook
→ Auto follow-ups that stop the moment someone replies (thread-ID matching, not just a timer)
One decision I'm glad I made early: I skipped LinkedIn scraping entirely. It violates their ToS and risks account bans — not worth it for marginal personalization gains when a company website gives you enough signal anyway.
Reply detection is built on Gmail thread-ID matching rather than reading email content — cleaner, more reliable, and it's the same mechanism you'd use in production. (Tested manually for this demo; the logic is identical either way.)
This isn't about replacing an SDR — it's about removing the manual bottleneck so a 2-person team can actually run outbound at all.
