The agent can take a while on a big job, and it tells you where it has got to.

​

​

The agent streams what it is doing as it goes, so you can see it working rather than staring at a blank screen. When it has finished, the reply is complete and the composer is ready for your next message.

​

When the agent works on a file, such as a spreadsheet or a long PDF, each piece of work shows as its own line with a green tick. Click the arrow on a line to see the code it ran and what came back.

​

A line marked failed means that attempt didn't work. Open it to see why, for example The analysis service was temporarily unavailable, so this step was skipped. The agent usually tries again on its own.

​

You can leave the page. The work carries on, and the conversation is in your chat history when you come back.

​