A dashboard from one sentence, built on a file you already have

What "generated" actually means here

Most tools that generate a dashboard from a prompt produce a mockup: a nice-looking arrangement of charts with sample numbers, which you then have to rebuild against real data. That is a design tool, and it is useful, but it is not a report.

What gets built here is wired to your tables. The AI decides which summaries the question needs, builds those summary tables in the workspace, points the charts at them and arranges the canvas. Change a number in the source table and the dashboard follows, because the charts are reading the table rather than holding a copy of it.

The steps

1. Get the data in. Excel, CSV, JSON, a database over JDBC, or an HTTP endpoint. Import is not capped on any tier. Importing.
2. Say what the report is for. Name the measures and the dimensions in plain language — the more specific the sentence, the less rearranging afterward.
3. Adjust it. Everything generated is an ordinary view: change the chart type, swap an axis, drop a panel, add one by hand. Nothing is locked because a model made it.
4. Keep it. Refresh the source data and the dashboard updates. On the paid tier, an automation can refresh from the source on a schedule so it is already current when you open it.

Sixteen view types, not four chart styles

A dashboard is only as good as the views under it. What is available: spreadsheet and sheet views, pivot tables, bar, line, pie, mixed, stacked and time-series charts, funnels, gauge and metric cards, forms, rich text and embedded web pages. The metric card and the pivot are the two that carry most real reports; the AI picks among all of them.

One dashboard is free

The free tier includes one dashboard, on purpose: a dashboard is the point at which this kind of tool either clicks or does not, and that should not be behind a card. Build a real one, put your own numbers in it, and decide from there. Additional dashboards, the AI generation itself, automations and uncapped export are on the paid tier. The split.

Where the data goes

TableDI is a desktop spreadsheet workspace that runs entirely on your machine. The tables and the dashboard live in a folder on your disk; there is no account and no sync. The exception is the AI call itself: generating the dashboard sends your prompt and a description of the relevant tables to the AI provider whose API key you entered, from your machine, directly. TableDI never sees the key or the data. Setting up a key.

What it will not do

• Publish it as a link. There is no share URL — the desktop app is single-user by design. You export an image or a PDF, or you show your screen.
• Guess a metric you never defined. Ask for "performance" and you get something generic. Ask for "gross margin by product line, monthly, last four quarters" and you get the report.
• Fix the data. A dashboard over a table with three spellings of one region name shows three regions. Clean first — how that works.