Task

Reply Suggestions from Pythia are not always relevant. How to improve them?

Resolution

During the trial, you get the basic level of Reply Suggestions from Pythia. There are different ways to improve them.

Macros titles and texts optimization
Pythia uses the text of the title and the comment of the macro for search and suggestions. You can define macros titles in a special way to categorize them and make it easier for Pythia to understand them better. Refer to this example of the categorization and titles for Ecommerce companies.
Custom ticket forms / fields as filters
Using custom ticket forms or field(s) to categorize tickets can also help Pythia to prefilter macros before making suggestions. This works well for tickets submitted via Help Center forms or Web Widget. For plain emails, you need to assign those forms / fields manually. Or you can use our Pythia Classifier to automate the categorization process (following the link, you can test it live!).
Advanced relevancy training
Using your historical data of macros usage we can train a custom AI model to improve the quality of the suggestions. For the frequently used macros, you will get usually a dramatic improvement in relevancy. Contact us for more details about this feature.

Shortcuts titles and texts optimization
Like with macros Pythia uses the text of the title and the comment of the shortcut for search and suggestions. You can do the same optimization with your shortcuts to improve the quality of suggestions.
Tags as filters
There are no forms or fields in Zendesk Chat. But the same result can be achieved by using tags. You can assign the tags to shortcuts related to different brands, languages, products, etc. Contact us for further assistance and setup of this feature.