The Engineering of AI Agents: Context, Harnessing, and Autonomy

Birgitta Böckeler from Thoughtworks explores the engineering practices behind building production-grade AI agents. The talk covers context engineering techniques like Anthropic Skills and MCP (Model Context Protocol), the concept of 'harness engineering' for building deterministic safety nets around non-deterministic models, and identifies the 'lethal trifecta' of AI security risks. Enterprise architects and senior developers learning to scale AI autonomy safely will find this particularly valuable.