Agents
Ask a question, get a dashboard you can keep
The Visualization Agent reads a plain-language question, writes the query against governed data, and returns a chart or dashboard that is correct, reusable and explainable.
What it does
The question becomes the query becomes the chart
Getting a chart usually means finding the right table, writing the query and choosing a visual. The Visualization Agent does all three from a question in plain language. It maps the question onto the governed data model, writes the query, picks a chart type that suits the answer, and shows you the query and the definitions it used — so the result is something you can trust and keep, not a black-box picture.
Language to query
Translates the question into a query against the semantic model, not a guess at raw tables.
Sensible visuals
Chooses a chart type that fits the shape of the answer rather than defaulting to one.
Shows its working
Displays the query and the definitions used, so the number can be checked and reused.
Inputs it accepts
What you feed it
Plain-language questions
A question in ordinary words, from one metric to a full dashboard request.
Governed data products
The published products and semantic model it is allowed to query.
Existing dashboards
A dashboard to extend or a metric to add, rather than starting from scratch.
Decisions it makes on its own
What it builds without asking
When the question is clear and the data is available, the agent builds the whole thing: it resolves the metrics against agreed definitions, applies the access rules of whoever asked, writes an efficient query, and renders an interactive result that can be saved and shared. Because it uses governed definitions, two people asking the same question get the same number.
Resolves definitions
Uses the agreed definition of each metric, so results are consistent across the business.
Respects access
Applies the asker's permissions, so a dashboard never reveals data they may not see.
Produces reusable output
Saves the result as an interactive, shareable dashboard rather than a one-off image.
What escalates to a human
Where it asks before guessing
Ambiguous questions
When a term could mean more than one metric, the agent asks which definition you meant rather than silently choosing one.
Missing data
If the question needs data that is not published or that the asker cannot access, it says so and points to what would be required, instead of returning a misleading chart.
Systems it connects to
Where it reads and writes
Data products
Queries governed products through the same access policies as any client.
Semantic model
Uses the shared definitions and relationships to resolve questions correctly.
Dashboards and reports
Publishes results into shared spaces where teams already look.
A worked example
Revenue by region, last quarter
Someone asks for revenue by region for the last quarter. The word revenue maps cleanly to the agreed metric, so the agent writes the query, applies the asker's regional access, and returns a bar chart broken down by region with the query on show. A second person asks the same question and, because both use the governed definition, they get the same totals — the argument about whose number is right never starts.
Questions
Frequently asked
- How do we know the number is right?
- The agent uses the governed definition of each metric and shows the query it ran. You can read exactly how the figure was produced rather than taking a chart on trust.
- Can people see data they should not?
- No. The agent applies the asker's access rules to every query, so a dashboard only ever shows what that person is entitled to see.
- Is the output a static image?
- No. It produces an interactive, reusable dashboard that can be saved, shared and extended, so good answers become shared assets rather than throwaway pictures.
Ask a question of your own data
Give us a real question and a governed dataset and we will show you the dashboard the agent builds, with its query on show.