Every visual here was built by an AI agent through Vilsem, previewed on live data and committed into a real Power BI report. Copy a prompt, point your agent at your own model, and adapt the fields. Install Vilsem · GitHub
Say what you want to learn first ("who is on track?"), then the chart, then the look. The agent picks better fields when it knows the question.
"Latest full year", "per month", "each quarter", "top 10". Ambiguous grain is the most common cause of wrong numbers.
Time order or ranked order. Years, months and overs stay in natural order only when you ask for it.
Dark card or light, gradient or solid bars, accent colours, labels on or off, a big year in the corner. Concrete words get concrete results.
"Preview on real data, show me, then commit." The agent sees the render, reads Vilsem's critique and fixes contrast, overlap and ordering itself.
Ask it to place the visual on the page without overlapping others and keep text readable, instead of forcing a tiny box.
"Make the labels bigger and the bars solid." Vilsem patches the stored spec with a before/after preview.
For Publish to web and the Build pane, ask the agent to keep measures in the model, not inside the visual (get_playbook "9a").
No builds match these filters. Clear a filter or the search.