Uploading business data

To upload your business knowledge, open the "Automation - Knowledge" link in the project menu and use "Add file" or "Add URL" buttons to upload as many documents as needed (please note that there's a 4 megabyte limit for each individual document file size).

Once you have this information in place, click the "Train the model" button in the top right corner. This will start the machine learning process, which usually takes from 2 to 24 hours to complete, depending on the volume and complexity of the data.

When the model is ready to use, you will get a notification email.

Responses given by that model can be used in three different modes:

  • the "total automation" mode, when every message sent by the customer is treated as a question, and every chatbot response is generated automatically by the AI (this is the default mode for every new project that you create in Activechat)

  • the "live chat hints" mode, when AI-generated responses are shown as "hints" to live chat agents during their conversations with customers

  • the "mixed" mode, when you're using the natural language understanding engine to detect customer's intent and then either run a specific scenario ("skill") for prioritised intents, or trigger an automated response for less specific and less important intents.

Updating the model with new data

Re-training the model when you add new documents to the knowledge base (or delete existing) is as simple as clicking the "Train the model" button again. Please note that until the training process is complete (you will get an email when it happens), your automated responses will still be using the old dataset.

The interface will show you which documents are already used by the model, and which require re-training. Also, the model status (shown in the top part of the screen) will change from "Active" to "Needs Updating".

Last updated