15 Months of building AI agents in 15 minutes
summary
Builds on 15 months of creating AI agents for Google Cloud customers and claims practical rules that work in production: use critical user journeys to tie every agent action to a business outcome; prefer deterministic code for tool-selection where possible and avoid needless multi-agent complexity, or error compounding that drops example success to 95%; test at atomic, integration and end‑to‑end levels; prefer binary pass/fail metrics and aim for at least 90% LLM‑to‑human alignment before trusting automated judges, track implicit user signals, and use meta‑prompting.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:42Product mindset for AI agents
- 3:30UX beyond chatbots
- 5:57KISS for AI agents
- 9:29Engineering AI agents
- 13:50Final thoughts & Q&A
related talks
27mHow to Build AI Agents and Keep Your Sanity
29mBuilding Trust in AI Agents
32mThe AI Agent Path to Prod: Building for Reliability
56mAI Agents That Don’t Suck: How to Build Ones That Actually Drive Business Value
16mBuilding Applications with AI Agents
49mHow to be an AI First company
15mPractical tactics to build reliable AI apps
07h 44mAI Agents Conference 2026 - Day 1