I built an outbound system where the prospect list is basically the easy part. The interesting part is everything that happens after the list exists. Here’s the workflow: 1. SIGNALS Instead of pulling 10,000 companies and hoping some are interested, the system watches for things like: → New funding → Hiring activity → Product launches → New executives → Expansion → Technology changes → Website changes Those signals become reasons to prioritize an account. 2. RESEARCH When an account gets flagged, an AI agent researches it. Company → product → ICP → recent changes → relevant people → likely pain. No generic “I saw your company is growing” research. 3. QUALIFICATION The system scores the account against the actual ICP. Good fit ≠ good opportunity. The goal is finding accounts that have fit + timing + a reason to talk. 4. PERSONALIZATION Only after qualification does the system generate outreach. The message is based on the signal and research not five variables merged into a fake-personalized paragraph. 5. ROUTING High-intent accounts get pushed to the right place automatically. CRM. Slack. Email sequence. Rep queue. Whatever the team’s workflow requires. 6. FEEDBACK Replies, meetings, conversions and rejection reasons flow back into the system. So the next batch isn’t built from the same assumptions as the last one. That’s the part I’m most interested in. Not: “AI can send cold emails.” We’ve known that. It’s building a GTM system that can continuously answer: Who should we target? Why now? What should we say? Who should handle it? What did we learn? That’s where GTM engineering gets interesting.

Imagine an outbound workflow that finds the right accounts, figures out why they’re worth contacting, and prepares relevant outreach automatically. Claude Code + APIs can make that possible. The interesting part is connecting the entire process.

the way i use AI for GTM: give it the goal. give it the context. let it research. let it find patterns. turn the good stuff into repeatable processes. keep feeding it feedback. it’s making the whole process easier.

My outbound setup is pretty simple: Pick the right ICP. Find a reason to reach out. Build the list. Test different angles. Keep the ones that get replies. The goal isn’t sending more cold emails. It’s finding what actually gets the right people to respond.

POV: You just started doing GTM work and suddenly everyone expects pipeline 😂 You’re supposed to figure out the ICP, build the list, research accounts, write the outreach, launch campaigns, book meetings… And somehow have results by the end of the quarter.

Cold email gets blamed for a lot of bad outbound. But if you’re sending the same message to 500 companies that don’t really need what you sell, the problem isn’t the email 😂 Your targeting is fucked.

Been digging deeper into GTM lately. Something I keep noticing: Most people focus on finding more leads. But the real leverage seems to be in how you research, qualify, and prioritize them before outreach. That’s the part I’ve been studying more.

Stop manually researching every prospect before outreach. Let the system handle the first pass: → ICP fit → Buying signals → Why now → Account score Then sales focuses on the accounts actually worth pursuing.

7 things I built into my AI outbound system: Research every account Check ICP fit Find relevant signals Score the opportunity Explain why it fits Generate the sales angle Route the right accounts It’s better leads for sales.

Building an AI workflow is the easy part. The hard part is making it actually useful. Knowing what data to collect. Which signals matter. Who actually fits the ICP. What deserves attention. That’s where the real GTM engineering happens.

Biggest mistake in outbound: Treating every prospect the same. A 10,000-account TAM needs volume. A 200-account TAM needs precision. Your GTM system needs to know the difference. Otherwise, you’re just burning sales time.

Verifying my @Engagio account: CG-ZA5Y #Engagio

being a small team changes which GTM systems make sense. you don’t need a 12-step autonomous outbound machine if you’re still figuring out who actually buys. at 500 target accounts, manually researching your best prospects might beat building a giant enrichment stack. at 50,000 accounts, automation becomes a different conversation. i’d rather build a simple system that helps me find the right 50 accounts, learn why they respond, and then automate what actually works. don’t engineer for a scale you haven’t earned yet.