← All news
Global · AI & Technology
Global·AI & Technology

Amazon Quick Smart Approvals automates review of eligible AI-agent actions

AWS describes how Smart Approvals in Amazon Quick checks proposed agent actions against user requests before approving them or seeking human review. Existing permission restrictions and mandatory human approvals remain in force.

AI Hubs United Editorial ·

Smart Approvals in Amazon Quick uses automated review to decide which eligible AI-agent actions can proceed and which need a person’s judgment, according to the AWS Machine Learning Blog. The feature checks proposed actions against the user’s request without expanding permissions or bypassing mandatory human approvals. Its practical aim is to reduce routine approval interruptions while preserving attention for decisions that need review.

A separate classifier model evaluates the action, its recipients and data destinations using limited conversation context. It can also consult working relationships in the user’s knowledge graph within the Quick account. AWS says reviewed content is not used for model training or service improvement. Existing controls determine eligibility before this assessment; blocked actions remain blocked.

Each automatic approval covers only the specific action and parameters assessed and can be used once. Timeouts, review failures and actions the classifier declines to approve go back to the user. AWS acknowledges that model judgment can be wrong and advises requiring human approval through connector permissions for sensitive actions. Smart Approvals is on by default, although deny-by-default profiles require explicit authorization. Users can disable it across conversations and scheduled runs, and account-level settings cannot override that choice.

AI Hubs United original short summary of the source linked above. It is not the original article; read the full story at the source.