Rate the agent's answers

Under the agent’s last answer of each Turn are a 👍 and a 👎. Use them to tell us whether the answer helped. You can rate any Turn that has ended, also one that failed.

Rate an answer

• Select 👍 (Helpful) when the answer helped. That’s all: the agent doesn’t reply to it.

• Select 👎 (Not helpful) when it didn’t. A dialog, What went wrong?, asks why. Both parts are optional:

  • pick the reasons that fit: Wrong numbers, Wrong data, Misunderstood me, Too slow or Other;
  • add a comment, for example what the answer should have said.

  Select Send to keep the 👎 with what you picked, or Skip to keep it without a reason. To take the 👎 back instead, close the dialog.

Your 👍 or 👎 shows the moment you select it. If it couldn’t be saved, the thumb goes back and a message says so; try again.

The agent follows up on a 👎

After a 👎, the agent replies in the Conversation: it asks what went wrong where your reasons don’t say it yet, and offers to fix the answer, or to remember how you want answers in future. A line, You rated an answer as not helpful, marks where the follow-up began; it isn’t a message of yours. Answer the agent as you would any message.

The agent doesn’t follow up while another Turn of the Conversation is still running, and never after a 👍.

Change or remove your Feedback

• To change it, select the other thumb. Switching to 👎 opens the dialog again.
• To remove it, select the thumb that is on.

Your Feedback stays with the Turn when you come back to the Conversation later.

Who can read a Turn you rated

Your Conversations are yours alone. Feedback is the one exception: a rated Turn, one you gave Feedback on, becomes readable for the users of your Tenant who hold the Right Read rated Turns, and for authorised staff of Scavenger AI, so we can see where the agent goes wrong and improve it. They see only the Turn you rated, never the rest of the Conversation.

When you remove your Feedback, or delete the Conversation, the Turn is private again at once.