AI handles anything a human can do at the computer, on their behalf

How to verify

Will a widespread AI agent authorization framework emerge that legally lets agents act on a user’s behalf (shopping, transfers, and similar)?