Virtual Assistant: Automatic Content Generation

This section of the user guide is focused on Automatic Content Generation. To help you navigate the article we have broken it down into the following sections:

• What is the Automatic Content Generation Feature?
• Who has access to Turn on Automatic Content Generation?
• How do I turn on Automatic Content Generation globally?
• How do I turn on Automatic Content Generation per Office?
• How does Automatic Content Generation work?
• How does Automatic Content Generation create responses?
• How do I know which questions were answered with Automatic Content Generation?
• How can I test the Automatic Content Generation feature before turning it on?
•  Why might Automatic Content Generation provide different responses to the same question?
• Updating the Content Security Policy to allow Automatic Content Generation

What is the Automatic Content Generation Feature?

Automatic Content Generation provides new generative AI auto content generation technology if turned on for the virtual assistant. When turned on it is globally used by the virtual assistant. When a user asks a question that is not already in the virtual assistant's knowledge base, generative AI uses harvested knowledge source data to automatically generate a response "on the fly" and presents the response in the virtual assistant. When turned on, Automatic Content Generation will be used for two way messaging as well. 

Who has access to Turn on Automatic Content Generation?

The feature is available to individuals with the following permissions: 

	Administrative	Standard	Limited	View Other Offices
Turn on

For more information on user permissions, review the User Roles & Permissions article.

How do I turn on Automatic Content Generation globally?

1. Under the Virtual Assistant selection, select Behavior Settings.

2. Select the Edit icon on the top right

3. Select Status On for Automatic Content Generation and then select Save

How do I turn on Automatic Content Generation per Office?

The Automatic Content Generation feature can be controlled at both the global and office levels, allowing for content customization to better suit the needs of each office. 

When activated, Automatic Content Generation provides personalized and relevant responses based on office-specific settings.

There are three setting options available: "On," which enables Automatic Content Generation for the office; "Off," which disables automatic content generation; and "Default/No Office," where the office inherits global settings unless manually changed.

When a student engages with the bot, the office is identified either by URL matching or through explicit office selection by the user. For two-way messaging, the office is determined by the office assigned to a campaign . For 1:1 text messaging, the office is determined by the reusable phone number linked to the office.

Permissions for updating Automatic Content Generation settings are granted to those with access to update Behavior Settings and editing rights for the specific office.

How does Automatic Content Generation work?

When a user asks the virtual assistant a question, the system will initially look for a response in the knowledge base. If no matches are found, then content will be automatically generated if possible. There can be several outcomes to the use of Automatic Content Generation:

1. Automatic Content Generation finds a response from the web knowledge source and provides a response.
2. If there are not enough matches in the web knowledge source to provide a response, then the virtual assistant will provide a communication opportunity as the response.
3. If the question asked is considered to be toxic, the virtual assistant will decline to answer the question and respond with "I'm here to provide helpful and respectful information. If you have any other questions, feel free to ask, and I'll do my best to help."

How does Automatic Content Generation create responses?

The following steps are used to create a response using Automatic Content Generation once the user inputs a question (query):

1. The raw query is analyzed using ChatGPT to internally generate multiple search queries

2. Relevant content is retrieved from the Knowledge Sources database

3. The relevant content is reranked. 

4. The ranked relevant content is sent to ChatGPT along with the original query to generate a response
   

How do I know which questions were answered with Automatic Content Generation?

Automatic Content Generation responses will not be saved in the knowledge base but will be retained and logged as "Generative AI" responses and marked with a GenAI" badge. To find interactions that used Automatic Content Generation access the interactions in the client admin.

For information on how to view Interactions, review the Virtual Assistant Transcripts & Interactions articles.

How can I test the Automatic Content Generation feature before turning it on?

Automatic Content Generation can be tested by enabling the Automatic Content Generation toggle within Test My Virtual Assistant. Automatic Content Generation does not have to be enabled within the Behavior Settings in order to test. 

For more information on testing Automatic Content Generation, review the Test your Virtual Assistant article

 Why might Automatic Content Generation provide different responses to the same question?

Our AI is designed to deliver the best possible answer based on the information it has at any given time. It learns from a wide variety of sources, so sometimes you'll see answers that vary slightly. This is expected functionality. It means the AI is continuously adapting to give you the most up-to-date and contextually relevant responses. 

One thing to keep in mind is that the way you phrase your question can also influence the response. AI takes the structure and tone of a prompt into account when generating answers. For example, a small shift in sentence construction or wording might lead the AI to emphasize different aspects of its knowledge. This flexibility allows it to be more nuanced, but it can also lead to variations. 

Updating the Content Security Policy to allow Automatic Content Generation