Agentic AI capabilities in Knowi are AI-powered assistants designed to take action within the platform. On top of answering 'why' and 'what' questions, agents can interact directly with Knowi features, such as translating natural language into dataset queries, creating or modifying visualizations, and exporting results, making it easier and faster for end users to explore insights. Knowi currently offers three agents: the Dashboard Agent, the Widget Agent, and the Query Agent.
Key Features
- Natural Language Interaction: Ask questions of your data in plain english and receive actionable recommendations and insights.
- Natural Language to Query: Describe what you want to know and the Query Agent automatically generates a query against the most relevant dataset, no SQL required.
- Automated Actions: Modify widget data, update dashboard style and settings, and export data without manual configuration.
- Guided Analysis: Receive AI-driven guidance to help interpret results and decide next steps.
How to Enable Agentic AI
Activation
Reach out to your Knowi account manager or email support@knowi.com to enable Agentic AI if it is not already available for your account.
Configuring AI Settings
Once Agentic AI is enabled for your account, navigate to User Settings from the bottom-left sidebar. Open the AI Settings tab and toggle on AI Agents Access if it is not already toggled on.
NOTE: This will control AI Agent functionality for your entire account. Disabling AI Agents Access will disable agentic functionality for all users under your account.
Role Permissions
Agentic AI usage is controllable through role permissions. From User Settings, navigate to the Roles tab. Under the AI category, enable the appropriate AI Agent Assistant permissions:
- Use AI Agent Assistant in dashboards – allows access to the dashboard agent
- Use AI Agent Assistant in queries – allows access to the query agent
- Use AI Agent Assistant in widgets – allows access to the widget agent
Only users with the relevant permission(s) toggled on will be able to access the corresponding AI agent. For more information on managing roles and permissions, refer to User Roles documentation.
Using the Dashboard Agent
The Dashboard Agent lets you ask questions and perform actions across a dashboard using natural language.
Dashboard Agent Capabilities
- Dashboard Recommendations: Receive AI-generated suggestions to help you better understand and explore your dashboard.
- Dashboard Filters: Create and apply filters that affect the entire dashboard.
- Dashboard Settings: Edit and manage dashboard-level configuration and preferences.
- Export Dashboard: Export the dashboard to PDF, PowerPoint, or Excel.
- Sharing Options: Generate share URLs, secure hash-based share URLs, or iframe embed codes.
Examples of Dashboard Agent capabilities are provided below.
Accessing the Dashboard Agent
To use the Dashboard Agent, click the Dashboard AI Assistant icon in the top-right corner of the dashboard.
This opens the Dashboard Agent panel, where you can interact with the dashboard using plain English. Use the chat to ask questions or request changes at the dashboard level.
Generate Dashboard-level Recommendations: The Dashboard Agent provides high-level insights and recommendations based on data across the entire dashboard. It helps surface patterns, comparisons, and notable changes across widgets. You can steer the analysis by asking broad questions or specifying areas of interest, such as overall trends or anomalies worth investigating.
- Example prompt: "which campaign is the most effective"
Apply Dashboard-level Filters: The Dashboard Agent allows you to create and apply filters that affect the entire dashboard at once. Use natural language requests to set, update, or remove filters across all widgets, making it easy to focus the dashboard on a specific time range, segment, or set of conditions without configuring each widget individually.
- Example prompt: "set customer filter to Target"
Share and Export: The Dashboard Agent lets you export the dashboard to Excel, PDF, or PowerPoint. When exporting to Excel, each widget's data is included as a separate tab. You can also generate a shareable URL, including secure or embedded options, to easily distribute or share the dashboard.
- Example prompt: "create a shareable url"
Using the Widget Agent
The Widget Agent allows you to interact with dashboard widgets using natural language. You can ask questions, adjust widget settings, generate recommendations, and export results, all without manually editing the widget.
Widget Agent Capabilities
- Ask Questions of your Data: Query and transform widget data using natural language.
- Customize Widgets: Update chart types and styling directly through prompts.
- AI Summaries and Recommendations: Receive AI-generated insights based on widget data.
- Export Data: Export widget results to CSV, Excel, or PDF.
Examples of Widget Agent capabilities are provided below.
Accessing the Widget Agent
To use the Widget Agent, click the AI Assistant icon in the top-right corner of the widget you want to work with.
This opens the Widget Agent panel, where you can interact with the widget using plain English. Simply type your question or request into the chat, and the agent will analyze the widget's data and take action directly on the widget.
Ask Questions of Your Data: Ask a question about your data and press Enter. The Widget Agent will analyze the dataset and apply the appropriate data transformation based on your request. The updated data will open in the Analyze tab so you can review and verify the steps the agent took to generate the result.
- Example prompt: "total monthly conversions by customer"
Customize Widgets: You can prompt the Widget Agent to update the widget’s chart type or styling. The agent will apply the changes and open the Visualization tab so you can review them.
- Example prompt: "create a stacked area chart of this data"
NOTE: Visualization changes are not auto-saved. You must click Save before exiting the widget to preserve any updates.
Get AI Summaries and Recommendations: The Widget Agent can generate AI-powered summaries and recommendations based on the widget’s data. You can guide the agent by providing specific prompts or focus areas to tailor the insights and recommendations it generates. Use prompts to highlight what you want the agent to analyze, such as trends, performance drivers, or areas for improvement.
- Example prompt: "what should I do to increase total conversions"
Export Data: You can export widget data by prompting the Widget Agent. Supported export formats include CSV, Excel, and PDF.
- Example prompt: "export to excel"
Using the Query Agent
The Query Agent lets you build queries using natural language directly within the Query Builder. Describe what you want to query in plain English, and the agent will connect to your datasources, configure the query, and populate the query editor so you can preview and run your data.
Query Agent Capabilities
- Natural Language to Query: Describe what you want to query in plain English and the agent configures it for you.
- Datasource Connection: Connect to and query any indexed datasource without manual configuration.
- Join Support: Request joins across multiple datasources in a single prompt.
Enabling the Query Agent
Before using the Query Agent, it must be enabled for each datasource you want it to access. Navigate to Queries & Data Sources and open the Data Sources tab. Click on a datasource to edit it. In the datasource settings panel, check Enable Query Agent to allow the agent to index and query that datasource.
Repeat this for each datasource you want the Query Agent to be able to access.
Accessing the Query Agent
To use the Query Agent, navigate to Queries and open a new or existing query. Click the Create with AI button in the top-right corner of the Query Builder.
This opens the AI Assistant panel on the right side of the Query Builder. When the panel first opens, the agent will begin indexing your enabled datasources. Once indexing is complete, you will see a confirmation showing the number of datasources and columns indexed.
Refreshing Your Schema
If your datasource schema has changed since the last index, you can update it by clicking Re-index schema at the top of the AI Assistant panel. This ensures the agent is working with the latest datasource structure.
Querying Your Data
Type your question or request into the text box at the bottom of the AI Assistant panel and click the red Submit button to send it.
- Example prompt: "query the full sendingActivity table from the mongo database"
The agent will interpret your request, identify the appropriate datasource and table, and populate the query editor automatically. If no dataset closely matches your question, the agent will fall back to your most recently used dataset. For the most accurate results, include the dataset name or relevant data terms in your prompt.
Once the query is ready, click Preview to load and review your data.
To reset the conversation and start a new request, click the grey circular arrows to the left of the text box.
Joining Datasources: The Query Agent supports joins across multiple datasources in a single prompt. Specify the tables and datasources you want to join and the agent will configure the join and populate the query editor.
- Example prompt: "join the full sendingActivity table from mongoDB with the full customer table from mySQL with a left outer join"
Name Your Dataset: You can prompt the Query Agent to name your dataset at any point during your session. The agent will update the dataset name field in the Settings section automatically.
- Example prompt: "name my dataset Customer Campaign Data"
NOTE: The Query Agent supports multi-turn conversations, so you can refine your query iteratively. After previewing results, you can follow up with additional requests in the same session, such as adding filters, adjusting joins, or renaming the dataset, without starting over.