Agent corrections
The AI does the first pass on every ticket, but you and your team have the final say. When a classification is wrong — the intent is off, or the detected language doesn’t match — an agent can correct it directly on the ticket. Corrections do two things at once: they fix the record in front of you now, and they teach the model to do better on future messages.
Making a correction
While viewing a ticket, an agent selects the correct value for the field the AI got wrong. The change applies to that ticket immediately, so any routing, filtering, or reporting that relies on the classification reflects the corrected value straight away.
Why corrections matter
- Accurate tickets today — routing and reports stay trustworthy because agents fix mistakes as they see them.
- A smarter model tomorrow — each correction is a labelled example of what the right answer should have been. Over time these examples improve the AI’s accuracy, so agents spend less time fixing and more time replying.
- A feedback loop your team owns — the people closest to your customers are the ones shaping how the AI understands them.
Getting the most from corrections
Encourage agents to correct classifications whenever they notice one is wrong, rather than working around it. A quick correction costs a few seconds and pays off across every similar ticket that follows. If you find the same category being mislabelled repeatedly, revisit your intent categories — the definitions or examples may need sharpening.