Zendesk CSAT Follow-up Agent

A CSAT dashboard tells you eleven people were unhappy last week and nothing whatsoever about why. This agent opens each low rating, reads the ticket the way a manager would, finds the exact message where it went wrong, checks whether that customer has been through this before, and leaves all of it on the ticket with one thing you could do today. It never messages the customer, because deciding to make it right is a human call and usually costs money.

What people use it for

• Every bad score gets read - Low ratings arrive faster than anybody reviews them, so most are never opened at all. Here each one is read within the hour it lands, on a Sunday as readily as a Tuesday.
• The moment it turned, quoted - The note carries the actual message where the tone changed, with its timestamp, so nobody has to reconstruct what happened from memory in a meeting a week later.
• Patterns, not incidents - The requester's history says whether this was one bad day or the third time this quarter. That is the difference between owing somebody an apology and a churn risk you can still catch.
• Fair material for coaching - Because it quotes the thread instead of characterising anyone, the note holds up in a one to one. Nobody is called slow; a fourteen hour queue is described as a fourteen hour queue.

Before you fork

What counts as low, and what happens to the good scores?

You set both in the follow-up rules. Anything you call good is counted and dropped without a note or a message, so a happy week is quiet rather than a channel full of congratulations. Only the scores you name get the full read, the internal note and the line to your CX lead.

Is this going to be used against my agents?

It quotes the ticket rather than characterising anyone, and it is instructed never to blame a named person for something the thread does not show. Notes sit on the ticket where your team can already read them, so nothing is written about somebody somewhere they cannot see it. In practice it defends agents as often as not, because the turning point is frequently a queue rather than a bad reply.

What does it need in Zendesk, and what does it cost?

Satisfaction ratings have to be switched on in Zendesk, and the connection needs to read tickets and comments, list a requester's other tickets, and add internal comments. Cost is one agent run per rating you choose to work, so it follows your low score volume rather than your ticket volume. Good ratings finish in a fraction of a run, because it stops as soon as it has checked the score.