NetApp puts approval gates around Workload Factory AI agents

NetApp has added AI agents to Workload Factory that can analyze storage environments, but they can make supported changes only after a user approves the action. The company introduced the agents in a September 30 administration update, and users can access them through a new Explore Agents section in its Ask Me interface.

The agents operate in read-only mode by default. They can inspect supported resources, identify risks, recommend changes and troubleshoot, but they cannot modify resources. Administrators can opt into read-write mode, though NetApp says each supported action must be explained and approved before an agent proceeds in the NetApp Console or Slack experiences.

Once read-write mode is enabled, an agent presents a summary and waits for approval before acting. NetApp’s release notes say approved actions can make changes to AWS, ONTAP or supported APIs, although the published pages do not provide a complete list of supported write operations.

Administrators can choose none, one, some or all configured AWS credentials for agent access. Choosing a credential does not add permissions: access remains limited by the current Workload Factory scope, the selected credentials and the IAM permissions already assigned to them. Without selected credentials, agents can answer general questions but cannot analyze the customer’s resources. NetApp says one agent can work across multiple AWS accounts using multiple credentials.

Optional settings expose more detail about agent reasoning and the tools and APIs used during analysis, providing a form of AI agent observability. Both verbosity options are off by default. NetApp says the agents support Amazon FSx for NetApp ONTAP and Google Cloud NetApp Volumes environments. It also says customer data is not used to train AI models. The opened sources do not disclose separate pricing, licensing requirements, regional limits or a staged rollout schedule.