Manage AI access, configure NLP search behavior, and select the AI models used across Knowi's AI-powered features.
AI Agents
- AI Agents Access - Enable or disable AI Agents for the account.
AI Model Keys
- Knowi In-House AI - Knowi's built-in AI model, enabled by default.
- OpenAI - Add or update your OpenAI API key to make OpenAI models available across AI features.
- Anthropic (Claude) - Add your Anthropic API key to make Claude models available.
- Google (Gemini) - Add your Google API key to make Gemini models available.
MCP Access Token
- Claude / MCP Token - Generate a token for use with Claude Desktop or other MCP-compatible clients.
AI Search / NLP
Controls NLP-powered search behavior and dataset indexing.
AI Search Settings
- AI Q&A Across Datasets - Enable natural language Q&A over your datasets. When enabled, the Index by default option automatically indexes new datasets for AI search.
- Bot Integration (Slack / Teams) - Enable AI Q&A via Slack or Microsoft Teams bot integrations.
- Saved Questions / Favorites - Manage a list of saved questions surfaced as suggestions in the AI search interface.
- Field Synonyms - Define synonyms for dataset field names to improve natural language query matching. See Field Synonyms for guidance on when to use synonyms vs. value aliases.
- Show Visualization First - When enabled, AI search results display a chart before the data table.
Global Alias Datasets
Define reference datasets used to map common terms and aliases to the exact values stored in your data. This helps Agentic BI resolve natural language queries accurately across all your datasets. To learn more, refer to Value Aliasing.
Global AI Instructions
Enter account-wide instructions that are applied whenever users ask questions of their data through Agentic BI. Use this to set context that should always be considered, such as business rules or reporting conventions.
Example: Fiscal year starts April 1. Always report revenue in USD. Never aggregate net sales across regions.
For authoring guidance on what makes a good instruction (and what to leave out), see Global AI Instructions in the Agentic BI docs.
Dataset Indexing
Lists all datasets in the account with two toggles per dataset:
- Indexed - Include the dataset in AI/Agentic BI so users can query it through AI chat. Only indexed datasets can be set as default. See Indexing for what indexing controls and when to leave a dataset out.
- Agentic BI Default - Pre-select the dataset for all users when they open the Agentic BI chat. Defaults apply customer-wide, but users only see defaults they have access to. Disabling indexing on a dataset automatically clears its default flag. See Default Datasets for guidance.
For the full setup workflow - including field naming, descriptions, direct vs. non-direct datasets, and runtime token design - see Setting Up Datasets for Agentic BI.
The list sorts indexed datasets first, then alphabetically by name. Use the search box to filter by dataset name. Click the pencil icon next to a dataset to open its dataset management page. Use Enable All or Disable All to bulk-update indexing across all datasets.
After making changes, a Save Changes button will appear at the bottom of the section. Changes are not applied until you click Save Changes.
AI Model Selection
Controls which AI model is used for each AI-powered feature. By default, all features use Knowi AI. Click Change next to any feature to select a different model. Available options depend on which API keys have been configured under the Settings tab.
- Agentic BI - Model used for AI agent workflows.
- AI Search - Model used for natural language search queries.
- Documents AI - Model used for document analysis features.
- InstantSights AI - Model used for automated chart and insight generation.
- AI Dashboard Generation - Model used for AI-generated dashboard creation.
- Recommendation Widget - Model used for AI recommendation widgets. Supports multiple models simultaneously (e.g., Knowi AI, GPT-4.1, GPT-4o, GPT-5, GPT-5 Chat).