SersweAI Outbound Operations Engine
Overview
The internal operating system that runs SersweAI's outbound: a Supabase-backed CRM with a 5-dimension lead audit and scoring engine producing A-F grades, a 4x-daily email send cron with pre-send MX verification and quota management, 30-minute IMAP reply polling with Claude Haiku intent classification, and a hot-lead detection layer that auto-generates a Telegram-delivered meeting prep brief (audit data, opening line, recommended package, case study, proposal outline). Daily 6 AM PT lead generation rotation across 64 unique neighborhood and category combinations. Healthchecks.io dead-man switch on every scheduled job. Diagnosed and fixed a silent reply-bounce that had eaten 7 weeks of prospect replies (DNS misconfiguration plus stale Gmail credentials), recovering 81 hot leads.
The Problem
Solo outbound at scale requires reliability discipline that most teams skip until something breaks. Without monitoring, MX verification, dead-man switches, and reply triage, replies disappear, deliverability craters, and hot leads cool down before anyone notices.
My Approach
Built a production CRM with end-to-end observability. The 5-dimension lead audit (Lead Response, Online Presence, Reputation, Marketing Automation, Operations) drives A-F grading. A and B grades trigger AI-personalized openers using audit findings, C falls back to template, D is filtered out. The 4x-daily send cron runs MX verification before every email, respects daily quotas, and handles deliverability throttling. Reply detection polls IMAP every 30 minutes, classifies intent with Claude Haiku (interested, not interested, out of office, needs follow-up), and triggers Telegram alerts on hot replies with auto-generated meeting prep. Healthchecks.io dead-man switches catch silent failures before customers notice.