How it works
Ask. See. Explore.
Three verbs replace six steps. Here is what happens between typing a question and looking at the answer.
How it works
Four steps, and only one of them is yours
You bring the data and ask the question. Everything between those two things is the product's job.
- 01
Bring your data
Upload a CSV or Excel file. Connecting the warehouse you already use is coming soon.
DashboardGPT profiles every column on arrival — types, cardinality, date grains, likely measures — so it understands the shape of your business before you ask anything.
- 02
Ask questions
Ask in plain language. No SQL, no field pickers, no chart builder.
Questions are resolved against a semantic understanding of your dataset, not a keyword match, so "how did the enterprise tier do last quarter" resolves to the right measure, filter and time grain.
- 03
AI analyses
The query runs, the result is checked, and the right visualisation is chosen for it.
Analysis runs on Superset-powered query execution. The visualisation is picked from the shape of the answer — a trend gets a line, a comparison gets bars, a single number stays a single number.
- 04
Explore visually
Every answer lands on an infinite canvas you can rearrange, extend and follow up on.
Follow-ups carry context. Select a chart, ask "compare this with last year", and the next answer builds on what you already have instead of starting over.
Under the hood
What happens when you press Enter
Nothing here is hidden from you: each stage is reported in the assistant panel as the run progresses.
- 1
Your question
“How did the enterprise tier do last quarter?” — written the way you'd ask a colleague.
- 2
Semantic resolution
The question is resolved against a profile of your dataset: which column is the measure, which is the tier, which is the time grain.
- 3
Query execution
An aggregated query runs against your data through the Superset-powered analytics layer. Only the result comes back.
- 4
Visualisation choice
The form follows the answer: a trend gets a line, a comparison gets bars, a share gets a donut, one number stays one number.
- 5
The canvas
The result lands as a node you can move, restyle or build on — and it becomes context for your next question.
The problem
Most of the work in BI happens before you learn anything
Between having a question and having an answer sit six steps, at least two tools and usually somebody else's queue. By the time the answer arrives, the question has moved on.
Traditional BI
- 1
Find the data
Track down which table, warehouse or export actually holds the numbers.
- 2
Understand the schema
Work out how the joins, grains and naming conventions fit together.
- 3
Build the query
Write SQL, wait, discover a wrong join, write it again.
- 4
Configure the chart
Choose a visualisation, map fields to axes, fix the formatting.
- 5
Assemble a dashboard
Arrange tiles, wire up filters, publish, wait for review.
- 6
Interpret the result
Finally look at the number — and think of the next question.
DashboardGPT
- 1
Ask
Type the question the way you'd ask a colleague.
- 2
See
The answer arrives as a chart, a number or a written insight.
- 3
Explore
Ask the follow-up. The canvas keeps every step of your thinking.
AI analyst
Show revenue by month.
Here's the monthly revenue trend for the last two years.
Added Revenue trend
Break it down by region.
North America leads on volume, but APAC is growing fastest.
Added Revenue by region
Which region performed best?
Two answers, depending on what you mean by best — here's both.
Added Regional performance
Change this to a bar chart.
Conversational analytics
Follow-ups that know what you're pointing at
Select a chart and ask “compare this with last year”. You don't restate the measure, the filter or the time range — the conversation carries them.
- Context travels
- The selected node, the active filters and the conversation so far are all part of the next question.
- Answers are editable
- Ask for a different visualisation, or change it directly on the node. Either way the canvas stays in sync.
- Stateful, not one-shot
- Each answer builds on the last, so an investigation accumulates instead of resetting.
- Explanations when they help
- “What caused the decline?” returns the reason and the numbers behind it, not just another chart.
The fastest way to understand it is to use it.
The demo opens straight into the product — a real dataset, a real AI analyst, a real canvas. No account, no install.