Guide & Implementation Flow — from "I built a workflow" to "I run a revenue system." The same play the best operators use.
Around 2024, companies stopped hiring people to do manual operations and started hiring people to build the systems that replace them. That role has a name now: AI Automation Engineer.
1. Enrichment Pipeline — raw list → qualified, contactable record. Scrape, AI-summarize, score against ICP (with disqualifiers), then run email enrichment. ~90% ICP fit vs ~50% scraping raw.
2. Scoring Model — score on timing, not just fit: recent funding, hiring the role you sell to, a new executive, stack fit, growth. 3+ signals = the account jumps every queue.
3. Reply Routing — inbound reply → webhook → CRM. Positives to a named owner. Track time-to-first-response like it's losing you money, because it is.
4. Signal Detection — watch ATS feeds (Greenhouse/Lever/Ashby), SEC filings, funding, exec changes, ad-lib activity. Hear first, not tenth.
5. Reporting Layer — slice by segment and persona. The aggregate reply rate is your least useful number; per-list positive-reply on trigger segments is the one that pays.
Need: SQL · Python or JavaScript · API skills (auth, rate limits, pagination) · an automation platform (n8n / Make / Zapier) · an agent layer.
Don't need: front-end design · production deploys · whiteboard interviews. The whole bar is: can you build a working pipeline and keep it alive?
Write disqualifiers as carefully as positives. Require a written reason for every verdict — no silent skips. Forbid invention: return none if no signal. Test on 50 records (10 yes, 10 no, 30 gray). Cheap model for obvious rejects, frontier model on survivors.
Ops/SDR → first working systems → AI Automation Engineer → Senior (owns plays) → reports into Growth (variable comp) → Consulting (price on annual value) → Agency (leverage the library). The library is the compounding asset.