Feedback training

Use this feature to improve your assistant's responses based on real interactions with your customers. You need it when you notice the system doesn't respond exactly how you want.

How it works

The system learns from mistakes and suggestions. For example, if a customer asks for Saturday opening hours and the assistant says it's closed, but you know it's open from 9:00 to 13:00, you can report the error to correct future responses.

Feedback comes from several channels: - Direct ratings (thumbs up/down) in the control panel or the app. - Reports coming directly from the chat channel. - Analysis of call recordings. - Summaries sent via [../callbot/email-di-riepilogo.md].

What you need to do

1. Check conversations or recordings periodically.
2. Rate the assistant's response using the feedback buttons.
3. Report imprecise or incomplete responses.

What glacom® does

It analyzes every report received and produces a targeted correction. This process updates the knowledge base to prevent the error from happening again. You can see the details of these operations in [../errori-e-correzioni/index.md].

Limits

The system does not automatically correct the error at the exact moment the conversation happens: the correction occurs after the feedback analysis. If the correct information is not present in any of your sources, the assistant cannot invent it.

Frequent problems

• I gave negative feedback, when does the response change? → After the report analysis has been processed and applied.
• Can I correct a response while the customer is in chat? → No, feedback training happens afterwards to refine the service.

From the app

You can quickly rate assistant responses and report errors directly from your smartphone.