Try Operator on web errands before you submit anything

Cassie Kozyrkov tested Operator on several browser tasks, including paperwork, planning, and shopping, and found it was strongest when the job only needed a decent result.

Done with Operatorยท Intermediate

What they did
Cassie Kozyrkov used OpenAI's Operator, a browser agent in ChatGPT Pro, and recorded it while it handled a mix of online chores. She featured hotel booking, research, flights, legal paperwork, and infographic work in the video, then also tried kids' party planning, Amazon shopping, and finding a freelancer outside the cut. Some of those side tests were left out of the final edit because they did not go as well.
How it went
The party-planning run was only okay, and the Amazon attempt failed when Operator could not place items into the cart. The freelancer search also went poorly, which she said was due to her prompt. She found Operator most useful when an answer just had to be good enough.
Tips
Give it one clear task at a time, and check its work before anything is paid for or sent. It was also very slow, so expect long waits and do not rely on it for parallel jobs or repeated loops.

Try it yourself with Operator

Help me with this online task: [plan a party for [number] guests on [date] / gather information about [topic] / find [items] on [store]]. Work one step at a time and tell me what you did and where you got stuck. Before you pay, submit, send, delete, or buy anything, stop and show me the result for approval. End with a short recap of what you completed and what still needs my attention.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on decision.substack.com โ†—

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.