A water-utility team is deciding between Cloud Run and Agent Runtime for two different workloads: a stateless outage-lookup agent that answers each request independently, and a stateful multi-turn casework agent for housing-authority intake that must track context across a resident's conversation. How should the workload characteristics guide the choice?

Workload state is a key factor in runtime selection — a stateless request-response workload like an outage lookup has no need to persist context between calls and fits comfortably on a general-purpose serverless platform like Cloud Run, while a stateful, multi-turn casework agent needs to track conversation and case context across turns, which is exactly what a runtime purpose-built for agent sessions and memory is designed to manage more natively. Claiming both fit Cloud Run equally ignores that session and memory management across turns is a distinct capability, not something serverless compute inherently provides. Claiming both fit Agent Runtime equally overlooks that a simple stateless lookup doesn't need session or memory machinery and can be well served by a simpler runtime. Preferring Cloud Run for the stateful workload because of scaling conflates scaling capability with session-state management, which are separate concerns. Scope caveat: runtime choice also depends on cost, existing infrastructure, and team familiarity, so workload statefulness is one strong signal among several rather than the sole deciding factor. Operational check: map each agent's need to persist context across requests before choosing a runtime, and confirm the chosen runtime's session or memory support matches that need.