Research shows that users are more inclined to accept AI acting fully autonomously in low-risk, reversible operations, while preferring to retain final confirmation rights when matters of taste, amounts, or dietary restrictions are involved.