In today's episode, I chat with Leo, founder at Ascend AI, about building AI automation infrastructure with ROI as the actual design constraint—not just shipping chatbots and email sequences, but end-to-end workflows that automate judgment, not just tasks. He walks through two campaigns in detail: a B2B SaaS client selling to VP Engineering at mid-market tech companies, where the team built a Clay waterfall using GitHub activity signals—open source dependency issues, legacy migration hiring—to score companies from zero to a hundred on signal density, generate a pain hypothesis for each lead, and show up in the inbox with messaging that referenced what was actually breaking in their engineering stack rather than a generic title-based pitch; and a D2C skincare brand on Shopify struggling with repeat purchase rates, where Leo's team built a post-purchase automation using Make and an AI orchestration layer on GPT-4 to predict next likely purchase, segment customers into education, replenishment, and cross-sell tracks, and trigger personalized emails, UGC sequences, and SMS offers in sequence—pushing repeat purchase rate from 20% to 34% within 40 days, email and SMS revenue per recipient up 62%, and support tickets down 41%. The throughline across both is the same: find the friction point, put intelligence into it, close the loop. Leo's path here is one of the more winding ones—laser engineering degree, left the country, taught himself digital marketing and SEO while freelancing abroad, met a product manager at a cafe in Azerbaijan and fell in love with the discipline on the spot, spent two years in PM working with startups until a growth roadmap he built got featured on Reforge, transitioned into GTM engineering when Clay started gaining traction, and eventually pivoted into AI automation when no-code tools made it possible to bridge the two without a developer background. His prediction: GTM stacks will eventually AB test their own logic autonomously, the GTM engineer role will morph into a hybrid of PM, data engineer, growth marketer, and DevOps, and the most interesting near-term development is models like Claude Opus acting as GTM architects that plan and delegate to other models and human devs underneath. His advice: think from the business owner perspective, not the tool perspective—ask what decision or task is stuck, not what you can build with a given tool—pick one vertical, master one or two tools in it, share your work in public as you go, and fix high-friction points end to end before adding more nodes on top. Enjoy 🙂 (0:00) Introduction to Outbound Wizards (0:32) What Ascend AI Does: End-to-End AI Automation Infrastructure With ROI in Mind (2:25) B2B Campaign: GitHub Signal Scoring, Tech Debt Hypothesis, Waterfall Enrichment for VP Engineering Outreach (6:10) Why Signal Work Only Matters If It Shows Up in the Actual Email (8:15) D2C Campaign: Shopify Skincare Brand, Post-Purchase AI Layer, Repeat Purchase Rate From 20% to 34% (11:22) Automating Judgment, Not Just Tasks: The Core Philosophy Behind the Agency (12:34) Leo's Journey: Laser Engineering to Freelance SEO to PM to GTM Engineer to AI Agency Founder (18:05) Predictions: Self-Optimizing GTM Stacks, Hybrid Roles, and Claude Opus as GTM Architect (21:41) Advice: Fix What's Stuck, Pick One Vertical, Master One Stack, Share in Public 🔗 CONNECT WITH LEO 👥 LinkedIn 💻 Website 🔗 CONNECT WITH SAURAV 🎥 YouTube Channel 🐦 X (Twitter) 📸 Instagram 💻 Website 👥 LinkedIn📧 Email - saurabh@salesrobot.co 🙏 LEAVE A REVIEW If you enjoyed listening to the podcast, we'd love for you to leave a 5-star review on Apple Podcasts to help others discover the show :) 👋🏼 GET IN TOUCH You can also reach out with any feedback, ideas or thoughts about the lessons you've learned from the episodes.