AI Chart Generator: Auto-Generate Dashboards
Use AI to generate a complete set of chart suggestions from your table metadata in six steps. No manual field assignment required. Your raw data never leaves your database.
Most BI tools use AI to explain dashboards you have already built. Edilitics does the opposite: the AI builds the dashboard for you. Describe what you want to understand from your data, choose how much context to share, and Edilitics returns a set of chart suggestions with Rows, Columns, and chart types already assigned. Select the ones you want and they land on the dashboard canvas, pre-configured and ready to publish.
You need a connected integration with a table selected in the chart builder before starting. Auto-Generate reasons about your table metadata, not your row data. The quality of the suggestions depends directly on how well your columns are described. Check the AIR Score for your table before running for the first time.
The AIR Score
AIR stands for AI Readiness. It is a score from 0 to 100, graded A to F, that measures how prepared your table's metadata is for AI-assisted generation. It is shown in the column selector header in Step 2.
| Grade | Score | What it means |
|---|---|---|
| A | 90 to 100 | Metadata is comprehensive and user-validated. Suggestions will be highly relevant. |
| B | 80 to 89 | Most columns are documented. Suggestions will be strong. |
| C | 70 to 79 | Some columns lack descriptions or validation. Suggestions may miss nuance. |
| D | 60 to 69 | Significant metadata gaps. Suggestions may be generic or misaligned. |
| F | 0 to 59 | Minimal documentation. Suggestions will be low quality. |
The score combines two inputs:
- Data quality: completeness, distinct value counts, null rates, and value ranges across your columns
- Metadata validation: how many column descriptions have been reviewed and saved by a user
AI-generated descriptions count for 20% of a column's contribution. User-validated descriptions count for 100%.
The fastest way to raise your AIR score is to open Step 2, read the AI-generated descriptions for your key columns, correct or complete any that are inaccurate, and click Save Changes. Each validated description immediately updates the score.
Auto-Generate works at any AIR score. A table with an F grade and no descriptions will produce generic charts that may not reflect your actual business questions. The same table with validated descriptions and a B grade will produce suggestions aligned to what your data actually measures.
How to Auto-Generate Charts
Open the Auto-Generate modal
In the chart builder, click the Auto-Generate button. A disclaimer modal appears describing what metadata may be shared with the AI provider you select.
Read the disclaimer. Click Use AI Assistance to proceed, or Standard Setup to close and build charts manually.
Check Do not show again to skip the disclaimer on future runs. What data is shared is always controlled by the privacy mode you select in Step 4.
Select columns and review metadata
The column selector shows every field in your table: categorical and datetime fields on the left, numerical fields on the right. Use the search box to find fields by name.
Select the columns you want the AI to reason about. The AI only uses the columns you select. Exclude anything irrelevant to your analytical goal. Fewer, well-described columns produce better suggestions than many poorly described ones.
Review and edit descriptions. Each column has a description that tells the AI what that field means in your business context. AI-generated descriptions are marked. User-validated descriptions carry more weight in the AIR score.
To edit a description: click the field, update the text (100 to 300 characters), and click Save Changes. The button shows the count of pending edits: Save Changes (3).
The AIR score is shown in the table header. See The AIR Score above for what it means and how to improve it.
Click Continue when your columns are selected and any edits are saved.
Describe your analytical goal
Enter what you want to understand from the data. This text anchors the AI's analytical narrative. Every chart in the suggestion set is oriented toward the goal you describe here.
Prop
Type
A character counter updates in real time. You cannot proceed with fewer than 50 characters or more than 300.
Choose a privacy mode and LLM provider
Control exactly what data context is sent to the AI, and which provider processes the request.
Maximum privacy. Column names are replaced with anonymous identifiers (col_1, col_2, etc.) before being sent to the AI. The AI receives the anonymized schema, data types, column descriptions, and data quality statistics.
Use Private Mode when your column names contain sensitive business identifiers or when your data governance policy restricts sharing schema details with external services. The AI can still generate valid chart suggestions from the anonymized schema and descriptions.
Sent to AI: Anonymized column names, data types, column descriptions, data quality statistics (null counts, distinct counts, value ranges), user role.
Never sent: Real column names, frequent values, row data.
Default. Real column and table names are shared with the AI alongside data types, column descriptions, and data quality statistics. Frequent values are not included.
Balanced Mode produces strong suggestions for most tables. The AI understands field semantics and data shape without seeing what values are actually in the table.
Sent to AI: Real column and table names, data types, column descriptions, data quality statistics, user role.
Never sent: Frequent values, row data.
Maximum reasoning power. In addition to everything in Balanced Mode, the AI receives the most frequent values for each column. This helps the AI understand categorical distributions, valid grouping values, and realistic filter ranges.
Use Full Context Mode when your table has categorical fields with non-obvious values (product codes, region identifiers, status flags) and you want suggestions that reference real categories rather than generic groupings.
Sent to AI: Real column and table names, data types, column descriptions, data quality statistics, frequent values per column, user role.
Never sent: Row data.
No privacy mode ever sends actual row data to the AI. Individual records never leave your database regardless of which mode you select.
LLM provider: Select the AI provider to process your request. Available providers depend on your plan. A provider must be selected to proceed.
Click Submit. If your AIR score or data quality grade is low, an advisory notice appears before the request is sent. Review it and confirm to continue.
Review and select chart suggestions
Edilitics presents between 5 and 15 chart suggestions. Each row in the table shows:
| Column | What it shows |
|---|---|
| Chart Type | The recommended chart type with an icon |
| Columns | Dimension fields assigned to the chart axis |
| Rows | Measure fields with aggregations assigned |
| Description | A business description of what the chart reveals |
Example. Using the sample orders dataset with the expectation: "Show revenue performance over time by region and product tier, and identify where growth is concentrated."
| Chart Type | Columns | Rows | Description |
|---|---|---|---|
| Basic Line | order_date (Group By, Month) | revenue (Sum) | Monthly revenue trend over 6 months. The baseline story of whether the business is growing. |
| Basic Bar | region (Group By) | revenue (Sum) | Total revenue by region. Establishes which markets are largest before going deeper. |
| Basic Bar | product_tier (Group By) | revenue (Sum) | Revenue by product tier. Identifies which tiers drive the most value across all regions. |
| Clustered Bar | region (Group By), product_tier (Group By) | revenue (Sum) | Revenue by region and product tier. Shows where tier performance concentrates geographically. |
| Scatter | region (Group By) | revenue (Sum), units_sold (Sum) | Revenue vs units sold by region. Tests whether high-revenue regions are also high-volume or just high-value per order. |
Notice how the suggestions form a story. The first chart establishes the overall trend. Each one after it adds a dimension of explanation. By chart 5 you have a complete picture of revenue concentration across time, region, and tier. This is by design: Edilitics generates suggestions as a set, not as isolated charts.
Select the charts you want using the checkboxes. Use the header checkbox to select all. Charts that match a visualization already on your dashboard are marked Already Added and cannot be re-selected.
Want more options? Click Re-Generate Charts. Edilitics generates a new set of suggestions and adds them to the top of the existing list. All suggestions from every run remain visible and selectable. You are choosing from a growing pool, not replacing the previous set.
You can regenerate up to 3 times per session, for a total of 4 generation runs. After the third regeneration the button is disabled. Click Start Over to return to the column selection step and begin a fresh session.
Click Create Charts when you have selected the charts you want.
Review each generated chart
Edilitics builds each selected chart in sequence on the canvas. A toast confirms the process has started:
Do not navigate away or make changes to the chart builder until all charts are complete. Interrupting the process may leave some charts unbuilt.
For each chart, Rows and Columns are pre-filled from the AI suggestion and the chart type is set automatically. The chart renders a live preview immediately.
Review the result. The chart is fully editable at this point: change aggregations, add or remove fields, switch chart types, or adjust formatting. When satisfied, click Add Sheet to commit it.
When all selected charts have been processed, Edilitics opens the dashboard canvas automatically.
FAQs
Next Steps
Build Your First Chart
Manual field assignment: connect a table, assign fields, and publish a dashboard step by step.
Aggregations
Full reference for aggregation options per field type and when to change the default.
Filters
Restrict chart data with hidden or interactive filters, including relative date conditions.
Dashboard Assembly
Canvas layout, sheet placement, notes, images, and canvas configuration.
Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.
Last updated on
Build Your First Chart
Connect a table, assign fields to Columns and Rows, and publish a governed dashboard in Edilitics Visualize. No code required. From integration to published chart in under 10 minutes.
Aggregations
Complete aggregation reference for Edilitics Visualize: numerical, categorical, and datetime options by field type, defaults, available formatters, and what belongs in Transform instead.