Clarifying questions

Tori asks the customer for more information when that information could change the answer. It doesn't normally guess which situation applies or choose the most common option without checking.

For example, setup steps may differ between two versions of your product. If the customer hasn't said which version they use, Tori can ask before giving the steps.

When Tori asks

If Tori can see that information is missing, it asks before researching. Sometimes it only discovers the missing detail while reading your sources. In that case, it stops to ask the customer before continuing with an answer that might not apply.

Tori uses the existing conversation before asking. If the customer has already supplied the information, it can use that instead of asking again.

What the customer receives

Tori asks a short, specific question. It can briefly explain why the answer matters.

It normally asks for the information it needs rather than sending a long answer covering every possible case. After asking, it waits for the customer to reply.

The question and the answer after the customer's reply are separate responses. Tori may do more research after receiving the new information.

Changing what Tori asks

You can tell Tori about details that matter for your product or your team's workflow. For example:

Before giving installation steps, ask which version of our product the customer is using, unless they have already told you.

You can also ask it to collect information that a human will need:

Before escalating a delivery question, ask for the order number if the customer hasn't provided it.

Collecting an order number does not give Tori access to the order. It gives your team information they can use when they take over the request.

Add these as custom instructions. To understand what carries over between messages, see Conversations and context.