AI is built on data. Review this table to find out which AI models Data 360 features support and the implications of turning off access to a model.
For more information on how to manage access to Large Language Models (LLM), see Manage Model Provider Access.
Note Salesforce uses data submitted to a predictive AI feature to train global
predictive models. Training these models improves your services and helps develop
new features available at no additional cost. The information we use includes
customer-submitted data, as defined in our Main Services Agreement (MSA), and usage
data related to user interactions with Salesforce.
| AI Features | AI Models Used | Implications of Disabling AI Models | More Details |
|---|---|---|---|
| Document AI |
Government Cloud: OpenAI GPT 4 Omni (OpenAI GPT-40) LLMs hosted on Microsoft Azure |
Creating Document AI configurations, accessing the testing playground, and processing any extractions (API or batch) for new or existing configurations aren't available. You have view-only access to activated document schemas. | Document AI |
| Notebook AI |
|
Notebook AI uses the GPT-4o model to process uploaded files and the Gemini model to generate answers. If you disable the model, the feature becomes unavailable. Users don’t have access to previously uploaded and processed data in their notebooks. | Explore and Research Content with Notebook AI |
| Intelligent Context |
|
Intelligent Context uses the model to analyze sample content to generate recommended processing configuration. It processes sample content by using LLM-based parsing and LLM-based pre-processing techniques. If you disable the model, the Intelligent Context feature is unavailable. | Intelligent Context in Data 360 |
| LLM-based Parsing and LLM-based Visual Data Processing |
|
LLM_based Parsing and LLM-based Visual Data Processing features aren’t available for a new search index and Intelligent Context configuration. The existing search index fails to process new content, but you can use previously processed content. | LLM-Based Parsing |
| LLM-based Image processing | Open AI GPT-4o | Use LLM to enhance responses from information available in images. If the model is disabled you can’t process new images using the LLM, but you can use previously processed content. | To enable image processing with a LLM when you select chunking strategies see: - Create a Vector Search Index with Advanced Setup - Create a Hybrid Search Index with Advanced Setup |
| Enriched Index | Claude Haiku 4.5 | The enriched indexing feature isn’t available for a new Search Index. The existing search index fails to process new content, but you can use previously processed content. | Enriched Indexing of Content Chunks |
| Agentic Segment Creation | OpenAI GPT4.1 | LLMs are used in Data 360 segmentation to enhance data insights and improve segmentation strategies.If you disable the model, segment creation fails. | Create Segments in Data 360 |
| Predictive AI Models | Models in Data 360 are global models. | N/A | Predictive AI in Data 360 |
| Predictive AI Sentiment and Classification Models | Models in Data 360 are global models. | N/A | Predictive AI in Data 360 |
| Salesforce Catalog | N/A | N/A | Salesforce Catalog |