The Everything Feed - All Packet Pushers Pods TCG068: Agents and Identity – Navigating What We Can’t Predict
Feb 11, 2026
Dan Moore, Senior Director of CIAM Strategy at FusionAuth and identity standards contributor, talks about how AI agents complicate customer identity, agent-specific authentication, and standards like OAuth/MCP. He outlines risks from non-deterministic agents and the need for agent modeling, finer-grained consent, guardrails, and practical ways practitioners can safely experiment.
AI Snips
Chapters
Transcript
Episode notes
Agents Add A New Identity Layer
- AI agents add a new intermediary between users and systems that greatly increases identity complexity.
- Ensuring entitlement and correct data flow to agentic requesters is a hard, critical problem.
Walk Before You Let Agents Act
- Start with guarded, advisor-style AI features rather than letting agents perform critical actions.
- Build small, instrumented helpers (e.g., doc assistants) before expanding to actor roles.
Doc Chatbot Delivered Real Value
- FusionAuth used RAG-based chat over documentation and saw big gains in surfacing information.
- Customers sometimes preferred asking the AI bot rather than contacting support.