Industrial AI Podcast Can AI Agents Really Run the Factory?
Mar 25, 2026
Tobias Ortmeier, robotics and automation expert and co-founder of Voraus Robotic, brings electrical engineering and medical-robotics roots to industrial automation. He discusses where generative AI and agents can help production, real-world pick-and-place trials, greenfield vs brownfield tradeoffs, and why human‑in‑the‑loop workflows currently outperform fully autonomous runs.
AI Snips
Chapters
Transcript
Episode notes
AI Agents Orchestrate Inputs And Actions
- AI agents act as orchestrators that collect inputs from multiple interfaces and coordinate retrieval, reasoning, and actions.
- Tobias describes the agent gathering data from maps, sensors and user inputs, then running retrieval and action steps to decide on maneuvers or code changes.
Agents Mitigate LLM Hallucinations With Retrieval And Tests
- LLMs generate token-based probabilistic outputs and can hallucinate with sparse domain data, so agents must fetch contextual data to validate suggestions.
- Tobias explains agents can retrieve internet or live production data and run tests to catch hallucinations before action.
Replace Proprietary Silos With High Level Code
- Move away from proprietary silos and use high-level languages across automation to unlock generative AI benefits.
- Tobias recommends centralised control and Python/Rust for drives, sensors and application code so LLMs can understand and modify systems.