AI Feature Overview
The platform has a number of features that are powered by generative AI models. Thrive’s current AI features include:
AI Feature | Description |
|---|---|
AI Assistant | A chatbot that gives users instant answers to user questions, grounded in customer content. |
Content Creator | A tool for course building, automated quiz creation and writing assistance tools. |
Content Summariser | A tool for generating concise summaries when viewing a piece of content. |
Tag Suggester | A tool for finding relevant tags based on content to improve discoverability. |
Tag Merger | A tool for merging similar/duplicate tags to keep content organised and consistent. |
🧠 AI Providers
Provider | Description |
|---|---|
OpenAI | Provider of models used for text and image generation. Models are inference via API. |
Google (Gemini) | Provider of models used for text, image and video generation. Models are inference via API. |
⚙️ Configuration
All AI features can be enabled by a Super Admin through the AI Configuration page in your Thrive dashboard. From there, the Super Admin can:
Provide consent for OpenAI and/or Google models to be used
Switch each feature on or off
Set whether features are available for all users or just admins
Data Privacy
📑 What we Share
OpenAI and Google have no direct access to data stored within Thrive.
The input to a generative AI (GenAI) model that outputs text, image or video is called a prompt, which is a set of instructions written in natural language. Each prompt consists of rules for how the model should generate its response, along with context that will help it do so. The context may consist of user inputs and/or customer content where relevant.
Thrive does not actively include personally identifiable information (PII) such as users’ names in prompts sent to GenAI models. We advise customers against including PII in their content as that content may be added to prompts for context.
💾 Data Processing and Retention
OpenAI and Google do not use any inputs/outputs to train future models. Thrive similarly does not train any GenAI models on these inputs/outputs.
Provider | Description |
|---|---|
OpenAI | Abuse monitoring logs, which may contain prompts, responses, and , are retained for up to 30 days. After that, the data is stored separately in a secure system indefinitely. It’s protected under legal hold, meaning it can’t be accessed or used for purposes other than meeting legal obligations. Only a small, audited OpenAI legal and security team would be able to access this data as necessary to comply with our legal obligations. OpenAI API requests are processed in the United States.
|
Abuse monitoring logs, which may contain prompts, responses, and metadata, are retained for up to 55 days before being securely deleted. Gemini API requests are processed in the United States.
|
🔌 Input Data by Feature
Feature | Customer data included in prompt | Reason for inclusion |
|---|---|---|
AI Assistant | User skills, goals and topics | Allows the assistant to provide more personalised responses |
AI Assistant | Content the user is viewing | Necessary for the assistant to answer user questions |
AI Assistant | Content retrieved via search | Necessary for the assistant to answer user questions |
Content Creator | User inputs specifying what should be included in the generated content | Necessary for the creator to generate intended content |
Content Summariser | Content the user is viewing | Necessary for the summariser to generate a relevant summary |
Tag Suggester | The content that tags are being requested for | Necessary for the suggester to provide accurate suggestions |
Tag Merger | A list of all existing tags | Necessary for the merger to identify similar tags |