About this talk
This talk addresses the critical issue of instruction compliance in AI agents, which often goes unnoticed as these agents become integral to software workflows. The speaker argues that while the industry tends to view this challenge as a context engineering problem, it is fundamentally about measurement and compliance. Drawing on research from 2025 to 2026, the session analyzes the phenomenon of instruction drift and why current practices struggle to detect it. The speaker also introduces a practical framework aimed at helping engineering teams determine whether their agents adhere to defined constraints and offers strategies for enhancing reliability in production systems.
More from this event
See all 126 talks →
AI Is Not the Risk. Architectural Drift Is - Sunil Kalkunte
17:39
Breaking the Monolith: Tesco’s Journey to Federated GraphQL with xAPI - Vishwas Chandrashekar
29:13
A Practical Introduction to LangChain4j - Venkat Subramaniam
1:01:28
Beyond the AI Models: How Lowe’s is Building the Store That Knows - Swaroop Shivaram
13:59