Capabilities / AI and conversations

Explore and build with Data Chat

Ask questions, build native and HTML pages, edit selected visuals, use voice input, and manage conversations.

X4 Tech SolutionsUpdated September 10, 20264 min read

Use Data Chat to ask questions, investigate results, and build or edit the report. One conversation can work with native visuals, HTML pages, shared model definitions, and report reference files.

A selected visual gives the assistant context. It does not create a separate assistant or prevent the conversation from working elsewhere in the same authorized report.

Start a conversation

  1. Open a report and select a page.
  2. Click Data Chat, then New chat.
  3. Type your request, attach relevant files, and select visuals if needed.
  4. Press Enter to send. Use Shift+Enter for a new line.
  5. Follow the activity and inspect the result on the report.
Example prompt
Inspect the model and explain the main drivers of the change in revenue.
Use the existing revenue definition. Show the supporting query results,
then create a new report page only after the analysis is clear.

Creating chats and running report-building work require the appropriate editing access. A viewer's ability to open a report does not imply permission to mutate it.

Choose the kind of result

RequestTypical result
Explain a metric or trendA written answer grounded in inspected context and query results.
Explore an ideaAn answer or inline preview in the conversation.
Create a chart or table on a pageA saved native visual with query, bindings, properties, and identity.
Build an HTML report pageSaved datasets, registered live visuals, layout source, and rendering checks.
Refine existing contentUpdates to identified objects, with saved-change receipts.
Change a reusable calculationAn inspected and updated model definition when supported and requested.

Say whether you want an exploratory answer or a saved page. An inline preview in chat and a visual saved on a report are different objects.

Work with selected visuals

Select one or multiple visuals before submitting a request. The selection context identifies the objects captured with that message.

Example prompt
For the selected visuals, use consistent date labels and number formats.
Keep their existing queries, data bindings, and filters. Include every
selected visual in your result summary.

The assistant can inspect current definitions before editing. HTML selections use registered visual identities; native selections use their saved native identities.

For a follow-up, name a new target when you change focus. Browsing to another page while a run is active does not retarget the submitted request.

Revise an inline preview

An inline preview belongs to a particular conversation message. The assistant can inspect that message and revise its preview in place.

Example prompt
Update the inline preview from your previous answer to show monthly
totals. Keep it in this conversation for now.

This does not automatically overwrite a similarly named canvas chart. Ask to save a report visual explicitly when that is the intended result.

Use reference files

Attach files in the composer or select existing ones from Report files. Uploaded files are available across conversations in that report.

The assistant can inspect document sections, search text, preview layouts, and reuse original embedded images. A file attached as reference does not become a SQL connection.

See Reference files for formats, limits, and examples.

Use voice input

Click Start voice input when the microphone control is available, allow microphone access, and dictate your request. Use Stop voice input to end recording or listening.

Recording-based input shows a transcription state before adding text to the draft. Review the recognized text before sending, especially identifiers, dates, and numbers. Voice input is a way to compose a message; it does not by itself submit a report change.

Support depends on browser microphone and speech capabilities. The status area explains permission or transcription errors.

Manage conversations

Use the conversation list to switch chats, rename them, or delete them. Names help separate investigations such as “Weekly review” and “Template redesign.”

Conversations are associated with their account and report. Sharing report objects does not automatically publish every person's chat transcript to all collaborators.

Deleting or closing a conversation is not the same as undoing its saved report changes. Use Change history to inspect supported edits.

Continue while work runs

You can draft the next message while a run is active. A conversation accepts one active request at a time. Use another conversation for separate work, and give concurrent edits distinct targets.

Navigation and minimizing chat do not cancel accepted background work. Use Stop to request cancellation. Already saved changes may remain.

Review an answer

Inspect the tool activity, effective SQL, and saved result. For HTML, confirm the assistant inspected a rendered preview after the relevant changes.

If an answer is incomplete, continue from the saved state and the reported error. AI context explains what the assistant can retrieve; Background work explains execution and cancellation.

Questions about your environment? Contact us.