Building an AI-Native Outbound Engine

SaaStr surveyed companies that had implemented AI SDRs and found only 3% had any revenue to show for it. Eddie's read is that this is the same pattern as Outreach and Salesloft a few years ago, and Gmail and Salesforce before that. So we mapped the system that has to be built around an ai-powered outbound motion before it will actually work. In this episode, Eddie Reynolds and Gerry Marletta, SVP of Revenue Operations Systems at USC, walk through how to build an AI-native outbound engine, agent by agent. You'll hear how they got around the catch-22 of deciding which accounts to enrich before you've enriched them, when to hand a step to an LLM and when to use an orchestration tool instead, what the agent that grades every rep email actually checks for, and the updates Gerry says a machine should never be allowed to make. Resources Mentioned in This Episode: Building an AI-Native Outbound Engine (Full Written Article) The AI-Driven Continuously Self-Improving ICP Free Outbound Pipeline Inspection USC Frameworks 04:18 - Only 3% of AI SDR buyers saw any revenue 08:18 - Why reps get 2,000 accounts and work 200 10:44 - The targeting agent and the data source waterfall 16:17 - Triggers and signals, and why one alone isn't enough 20:19 - The tool stack and why Salesforce stays the source of truth 23:33 - Why separate agents beat one big one 26:00 - When to use an LLM and when to use n8n 30:15 - The catch-22 of what to enrich first 39:50 - Scoring: start with a hypothesis and iterate 47:55 - What a good sales email actually looks like 56:57 - Who gets a fully personalized message and who doesn't 01:03:59 - The updates a machine should never make 01:04:36 - The first question for your AI analyst: did we execute 01:12:35 - The GTM council meeting that forces the loop _________________________________________________________ GTM STRATEGY & AI ENGINEERING FOR B2B TECH ● Website ● LinkedIn ● TikTok