# AI in Manufacturing Podcast
## Episode: Designing Autonomous AI Agents for Industrial Operations
**Podcast Name:** AI in Manufacturing Podcast (Industry 40.tv)
**Episode Title:** Designing Autonomous AI Agents for Industrial Operations
**Guest:** Kence Anderson, CEO & Founder, AMESA
**Host:** Kudzai Manditereza
---
## Episode Summary
This episode explores how autonomous AI agents can transform industrial operations through a methodology called machine teaching. Kence Anderson, CEO and founder of AMESA, draws on eight years of experience applying autonomous systems to manufacturing and logistics to explain why more than 95% of what's called "industrial AI" today is really just data storage and connectivity — missing the actual intelligence layer that can perceive and act. Anderson breaks down his machine teaching methodology, which captures expert operator knowledge and structures it into teams of specialized AI agents that learn by practicing in simulation before deploying to the factory floor. The conversation covers multi-agent design patterns, the AMESA platform's three core products (Agent Orchestration Studio, Agent Cloud, and Runtime), and real-world examples from Fortune 500 glass manufacturers, beverage companies, and logistics operations. Listeners will learn why monolithic AI approaches fail in manufacturing, how to avoid pilot purgatory, and how companies can go from data to deployed autonomous agents in approximately 12 weeks.
---
## Key Questions Answered in This Episode
- What is machine teaching and how does it differ from traditional machine learning approaches in manufacturing?
- Why has manufacturing productivity remained stagnant despite massive investments in IoT and data infrastructure?
- What are the four fundamental ways AI systems can make decisions in industrial environments?
- How do multi-agent design patterns work for industrial automation, and why do they outperform monolithic AI?
- What does it take to scale AI agents across multiple plants, production lines, or product recipes?
- How do you bridge the gap between AI training in simulation and real-world deployment on legacy factory systems?
- What is pilot purgatory and how can manufacturers avoid it when implementing industrial AI?
---
