Governing agentic AI inside internal engineering systems

This session explores how agentic AI is being embedded into internal engineering systems and why governance becomes critical as these systems gain autonomy. While agentic workflows promise faster resolution of engineering tasks, poorly designed autonomy can introduce new risks, failure modes, and loss of control.

In this talk, Rianat Abbas examines what happens when AI systems are allowed to initiate actions, approve changes, or resolve tasks inside internal engineering environments. The session highlights where agentic systems deliver value, where they break down, and why governance must be designed intentionally.

The talk covers:
• How agentic behavior reshapes risk, ownership, and accountability
• Common failure modes teams encounter as autonomy scales
• Where human-in-the-loop controls are essential to maintain trust and prevent silent errors
• Practical governance patterns that preserve developer velocity without compromising safety