- Category: Automation and AI, ITSM
- Posted by Jon Stevens-Hall
- on
Agentic AI systems build responses nobody specified in advance, and traditional governance models weren’t built for that. Jon Stevens-Hall draws on complexity thinking, from Cynefin to LLM-as-judge testing, to ask what it takes to govern behavior that develops rather than behavior that’s designed, and why loosening control in the right places might be the only way to keep it.