Plan requirement
| Subscription | Any plan |
| Access | Admin |
You pay when the AI agent resolves something on its own. What counts, what does not, and the questions to settle first.
The principle
Zendesk describes it plainly: you pay for requests the AI agent resolved successfully, without escalation to a human agent. A conversation the AI could not handle, which goes to a person, is not an automated resolution.
That alignment is the appeal. You are not paying for a bot to fail politely.
Where it gets less simple
The interesting cases are at the edges, and they are the ones to establish before you switch anything on.
- A customer who comes back an hour later about the same thing. One resolution or two?
- A conversation that ends without escalating because the customer gave up rather than because they were helped.
- A follow-up ticket arising from something the AI answered.
- Whether an allowance is included in your plan before charging begins.
None of these are trick questions, and Zendesk documents its definitions. They are worth reading rather than assuming, because they decide what your bill does as volume grows.
Watch the number from the start
Automated resolutions are reportable. Put them on a dashboard from the first week, next to your ticket volume and your escalation rate, so you can see the three move together. A rising resolution count with a flat ticket volume means something different from a rising count with rising volume.
The good problem
If your AI agent works, this line grows. That is what success looks like, and it still has to be budgeted for. Teams that model AI as a saving and not as a cost are the ones surprised in month four.
Comments
0 comments
Article is closed for comments.