Guardrails in Practice: Grant AI Agents Autonomy Without Losing Control
Guardrails for Your AI Agents
AI agents are moving from experiment to enterprise, but autonomy without boundaries is an overlooked liability. In this session, Micheal Lanham, author of AI Agents in Action, 2nd edition, breaks down what guardrails actually are in production agentic systems and why getting them right is the difference between a trusted AI workflow and an unpredictable one. Micheal walks through the layers of guardrail design: from input/output validation and typed handoffs to using agents themselves as validators in multi-agent pipelines. This session is designed for technical leaders, platform engineers, and AI architects building or evaluating production-grade agentic systems.
- Why non-determinism in LLMs makes external guardrail design non-negotiable in production systems
- How to implement input and output validation at the agent and agent-to-agent level
- Patterns for embedding guardrails in multi-agent handoffs without sacrificing performance
- How environment-level governance connects to agent-level control across the full AI development lifecycle
Meet the speakers
Micheal Lanham
Bestselling author and innovator
Micheal Lanham is a distinguished software and technology innovator. He is author of AI Agents in Action, Evolutionary Deep Learning, and Augmented Reality Game Development.
Jess Haberman
Content Strategy Director, Anaconda
Jess Haberman collaborates with tech industry leaders to develop instructional content in data and AI. She has presented at and facilitated technology conferences, webinars, live training courses, podcasts, publishing seminars, and writing retreats.