Description

“LLM agents are now good enough to do real work yet the most successful 2025 deployments still keep humans in the driver’s seat. In this session we will learn end to end co-creation loops where AI agents propose, humans tweak, and the system learns in real time. A small overview on co-creation loops: 1. a planner agent decomposes work; worker agents call tools 2. a policy layer computes confidence score × risk level and 3. domain experts inject clarifications or edits through a streamed, diff-aware UI. Corrections flow back as structured feedback, prompt patches, and vector memories, so the system improves after every interaction. Attendees learns design checklists, and a mental model for when to automate with agents, when to escalate, and how to turn every human correction into training AI agents. Whether you’re a product owner, engineer, or compliance lead, you’ll be ready to launch agentic AI teammates that speed up work.”