NVIDIA’s Jacob Liberman on the Power of Agentic AI in the Enterprise - Ep. 250

What It Covers

Jacob Lieberman, NVIDIA Director of Product Management, explains how agentic AI enables large language models to reason, act, and execute tasks autonomously in enterprise environments, transforming workflows beyond simple chatbots.

Key Questions Answered

• •Agent Evolution: Agentic AI represents the third era of GenAI use, moving from chat interfaces to retrieval augmented generation to autonomous systems that reason, plan, and execute tasks like booking trips based on preferences without human intervention.
• •Token Economics: The majority of future LLM-generated tokens will serve agent-to-agent communication rather than human interaction, similar to computational finance where 75-80% of stock trades occur between machines, fundamentally changing inference workload patterns and infrastructure requirements.
• •Autonomy Framework: Agent autonomy should map to risk levels—customer service agents need creative latitude with low risk exposure, while retirement portfolio agents require strict determinism. Enterprises can embed appropriate autonomy levels into agent actions based on this risk assessment.
• •Standardization Gap: Lack of standardization in agent communication protocols and memory storage creates friction when agents interact across platforms. This inefficiency drives up costs and prevents deterministic business outcomes, requiring checkpoint systems and unified frameworks for enterprise adoption.