AI agents are rapidly moving beyond developer prototypes into production enterprise environments, bringing new demands for security, compliance, scalability, and operational reliability. In a discussion with Daniel Whitenack and Chris Benson on Practical AI, Nick Kuhn from VMware Tanzu Platform outlined how organizations deploying agents at scale must adopt fundamentally different architectural approaches than traditional software deployment models. The shift requires enterprises to rethink platform infrastructure around several key capabilities: agent build packs that standardize deployment patterns, MCP gateways for managing external tool access, shared memory systems for state management, identity and access controls, and sandboxing to isolate agent execution. Kuhn drew on years of experience from platform engineering practices, showing how lessons from containerization, orchestration, and infrastructure-as-code can be applied to the unique challenges of agent-based applications. The discussion emphasized that deploying agents is not simply running new code on existing infrastructure—it demands new governance, monitoring, and architectural patterns.