How Chatty Leads Builds AI Agents Businesses Actually Want
summary
Most web chatbots fail because they either don't understand user intent or can't actually take the requested action. The build here stitches LLM conversation to real systems: a real-time email verification API called in-conversation, deterministic fallbacks, a waterfall of models for failover, and Playwright uptime checks. Early manual QA of 150–200 conversations a day drives an eval-based QA loop. Result: deployed agents drive measurable outcomes, including a reported 2–3x uplift in qualified leads.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:17Chatty Leads demo
- 8:41Four principles for agents
- 10:31Principle one: User needs
- 20:32Principle two: Outcomes over features
- 24:47Principle three: Owning client problems
- 27:18Principle four: Agent-centered systems
- 33:20Key takeaways and closing advice for builders and clients
- 36:12Q&A
related talks
19mThe Future of AI
35mDesigning Experiences for Chatbots
13mExploring AI-Powered Chatbots and Their Transformative Role
29mAvoiding the Chatbot Trap Designing AI Interfaces That Unlock Real Business Value
49mHow to be an AI First company
01h 04mBuilding Expertise Into Generated Conversations
56mAI Agents That Don’t Suck: How to Build Ones That Actually Drive Business Value
54mIntroduction to Chatbots