Features / Support
Find out whether that actually helped
Ask after a conversation, or ask a wider question whenever you want. Ratings attach to the conversation, the agent and the product, so the number always has a story behind it.
And the rating is filed against who resolved it
What it does
Without it
You closed it. You have no idea if it helped
The thread ends with 'let me know if that sorts it.' Nobody replies. You mark it resolved and file it under fine, because there is nothing else to file it under.
The score turns up with no story attached
A number lands in a slide once a quarter. It went down two points. Nobody can say which conversations, which person, or which part of the product moved it, so nothing changes and the next quarter looks the same.
Asking anything else means a second tool
You want to ask forty customers one question about pricing. That is a survey app, a list export, a mail merge, and a spreadsheet where the answers no longer connect to the person who gave them.
How it works
Pick the moment you ask
When a conversation resolves, when the AI handled it on its own, after the first reply, or only when an agent asks. A delay setting lets you wait rather than asking in the same breath as the answer.
What that includes
Pick the scale you actually want
Thumbs up and down, one to five, or NPS from zero to ten. Every score is mapped to positive, neutral or negative by one fixed rule, so three different scales can still sit on one chart.
What that includes
Each rating carries who handled it
Tore reads the resolution record and tags the rating: resolved by AI, by a person, or by a person working with AI. That comes from how the conversation was closed, not from anyone typing it in afterwards.
What that includes
Ask a wider question when you want to
Standalone surveys: one question, as text, a rating or multiple choice, placed in the widget, in an email, or on a link that needs no login to answer. A known customer, a signed-out visitor and an anonymous answer are stored differently, and the database allows one response per respondent per survey.
What that includes
Set it once, override where it matters
Satisfaction settings live at the organization level and any product can override them. The screen tells you which of the two you are looking at before you change anything.
What it means for you
The number has names behind it.
When satisfaction dips you open the conversations that caused it, in the same system, that afternoon. No export, no join, no waiting for someone with SQL.
You can tell whether the AI is helping
AI-handled and human-handled ratings are kept apart by default. 'Should we let it answer on its own' becomes a question with evidence rather than an argument between two opinions.
Asking your customers something stops being a project
One question, one link, nothing for them to sign into. The answers come back attached to the contact and the product instead of to a row in a spreadsheet nobody opens again.
The honest limit: Once a survey starts collecting, its question, type, answer options and scale are frozen. You can rename it, change the audience note or close it early, but if the wording was wrong you start a new survey rather than quietly editing the one people already answered.
Also in Support