AI-native describes software, services, infrastructure or organizations that are designed from the outset around AI models or agents, as opposed to existing products and systems that have AI features bolted on. The term is modeled on cloud-native, which denotes systems built on the assumption of running in the cloud.
An AI-native application treats model inference and agent behavior as its core functionality, while AI-native infrastructure refers to environments designed from the start to handle agent execution, permissions and isolation. It is widely used as a design philosophy or marketing term rather than a precise technical specification.