Predict Tech E-commerce Intent V1

Predict Tech E-commerce intents with hierarchy dependancy. Supported languages: en, fr

This action is used to predict the intent of an e-commerce customer request.

This includes tech products businesses. For apparel and perishable goods or subscription based businesses, please use the appropriate action.

To see the list of all supported intents, click here.

📘

The Hieararchy AI model

This action implements a hierarchy architecture that goes through 2 layers of AI Models. The first model will predict the general category of the request (returns, quality issue, miscellaneous, account...) and the second AI model will predict the intent, a more granular prediction.
Sometimes, the AI will not be confident enough to predict a category or an intent.

🚧

Preprocessing & Language Detection

the AI will automatically detect the language of the message and will clean the input customer message from any particular formatting (Email chains, HTML, URLs, special characters...) so you don't have to do it in the flow before using the action.

Inputs

InputTypeDescription
RegionStringServer region you can find in the url (us or eu)
DG Api KeyStringRead here
SecretStringRead here
MessageStringCustomer request

Outputs

IntentStringPredicted intent in the format "Category :: Intent", if AI was not confident enough to predict an intent, this field will display Category
CategoryStringPredicted Category
Intent FoundBooleanTrue / False
Category FoundBooleanTrue / False
LanguageStringPredicted language code, ISO 639-1 format (en, fr, nl...)
AI TextStringMessage after AI preprocessing
SuccessBooleanTrue / False, whether API call was successful or not
Status CodeIntegerREST API status code, success is 200
TagStringTag of the predicted intent that you can add on the ticket
Intent2StringThe AI too can be confused sometimes, this is a second high probability prediction. It doesn't appear often but you can use this one for messages with multiple requests for example
Tag2StringTag of intent2
InsightsObjectInsights of the prediction with preprocessing steps