Also available on:
In this episode of GydeBites, host Prasanna Vaidya speaks with Sudhanshu Hate about what it takes to build reliable Agentic AI systems in enterprises. Sudhanshu explains how agentic systems differ from traditional software, how agents use enterprise data through APIs and databases, and why most AI pilots struggle to reach production.
The conversation explores key production challenges, including security, scalability, session isolation, contextual personalization, and observability. Sudhanshu also shares practical insights on evaluating agent performance and on building AI systems that are safe, reliable, and ready for enterprise-scale deployment.
This episode is valuable for leaders and teams working to bring AI from pilots into business environments.
How is agentic AI fundamentally different from traditional systems?
How can agents leverage existing knowledge in the system?
How should organizations scale their Agentic AI initiatives and move from early pilots to production?
How do you evaluate Agents, and what metrics are used?
How much should one rely on the autonomous behavior of the Agentic System? What is needed to ensure this is safe and responsible?
Principal AI, Amazon Web Services
With over 27 years of experience building large-scale technology platforms, Sudhanshu has worked at the intersection of architecture, problem-solving, and execution. He enjoys tackling complex, first-time challenges and helping teams turn ideas into systems that work reliably at scale.
