Plan requirement
| Subscription | Suite Professional or higher |
| Also required | Copilot |
| Access | Admin |
What Copilot is doing across the team, and how to read it without turning a tool into a performance measure.
Look at it
- Open analytics and pick the prebuilt Zendesk Copilot: Agent productivity dashboard.
- Choose a tab: Auto assist, AI suggestions, or the generative tools.
- Filter by time and by ticket group, brand, form or channel. Agent name is a filter too, and the section below is about when not to use it.
- Compare against the period before.
The Auto assist tab leads with four figures: agents who accepted an auto assist at least once, tickets where one was accepted, applied suggestions, and the acceptance rate. Read that last one carefully: accepted or edited counts as accepted, so an agent who rewrites a suggestion entirely still raises it. The AI suggestions tab covers merge suggestions, quick answers, similar tickets, suggested first replies and suggested macros.
Clone the dashboard if you want to change it, and remember the figures follow the Explore refresh schedule rather than appearing the moment somebody accepts a suggestion.
The names to use if you build your own
Cloning the dashboard puts you in Explore, where the figures have names. In the Copilot auto assist dataset, % Acceptance rate counts interactions and % Suggestion acceptance rate counts suggestions, which is why the two can differ on the same period. % Dismissal rate is the mirror image and is often the more honest one to show beside it.
The question worth asking
Did handling time fall on the ticket types where Copilot is used, without quality falling? That is one comparison, and it answers whether the add-on is paying for itself.
Group level, not individual
Per-agent usage looks like a productivity measure and will be read as one, whatever you say. The moment it is, agents accept suggestions they would otherwise improve, and the reply quality falls while the numbers rise.
Group level answers the business question and avoids that entirely.
Low usage in one group
Usually means the work does not suit it, or the knowledge base does not cover that area. Both matter; neither is a problem with the agents.
Read quality alongside
Satisfaction, reopens, and repeat contacts on the same subject. Faster replies that come back a day later are not faster.
Look at the seats
Agents with access who never use it are the clearest number here. Either reassign the seat or find out what is in the way.
Report it in one line
Handling time on the affected ticket types, before and after, with quality flat. That is what a budget holder needs and it fits in a sentence.
Comments
0 comments
Article is closed for comments.