Shibumi utilizes OpenAI’s GPT-4, a large language model, for its AI Support, AI Analyze and AI Create Dashboard features. Large language models like GPT-4 are pre-trained on extensive datasets to learn language characteristics, enabling them to perform tasks such as language understanding and generation. Shibumi does not train models on client data and uses GPT-4 in its standard form as provided by OpenAI for handling language-based tasks.
For AI Support, the data provided is the Text Input (i.e., user entered questions) and the articles included on the Shibumi Online Support Site. No client program data is shared.
For AI Analyze, the data provided is the Text Input (i.e., user entered questions) and the pre-defined Shibumi Program work item Names and Score attribute values. No other program data is shared.
Note: AI Analyze is not available if the prerequisites are not satisfied (i.e., on a template named Initiative/Opportunity, the AI Score attributes must be defined). When the prerequisites are not satisfied, because AI Analyze is therefore not available, no Program data is shared.
For AI Create Dashboard, the data provided is the Text Input (i.e., user entered prompt) and the solution structure (e.g., template names and attributes). No program data is shared.
For AI Recommend, the data provided is the dynamically driven expression-based system prompt configured by the App Admin, this can include attribute references and their supporting definitions. The only program data shared is Work Item names.
Note: AI Recommend is not available if the App Admin has not enabled AI Create on the Template type displayed in the Table section, as well as configuring the Attribute Definitions and System Prompt. When the configuration prerequisites are not satisfied, because AI Recommend is therefore not available, no Program data is shared.