Ask and refine a question

Start with a question you care about. Querio works through it in an Explore, where you can inspect the notebook behind the answer and ask a follow-up when something needs clarifying.

Before you start

Open Querio and start New Explore. If you are new to the app, follow Get started in the app. To ask about company data, you need a connected datasource and permission to query it. Connect a datasource covers setup and access.

1. Choose the data needed

Open Select datasource beneath the new Explore prompt and select the sources relevant to your question. The control shows a selected source's name or a count when several are selected. Check it even if a source is already selected.

The model control shows the selected model or Auto. You can keep that choice or select another available model. If you change models in an existing chat, wait for the change to finish before sending. If you see Model change failed. Try selecting it again., resolve it first. Changing models does not rerun earlier results.

In an existing chat, the datasource icons beneath the prompt summarize its sources when you hover over them. To work with a different source, start New Explore and select it there. Selecting or mentioning a datasource does not grant query access; see Users, roles and groups.

2. Point to relevant context or cells

If a question depends on a particular table, definition or earlier result, point Querio to it. Type @ in chat, or use the Macros button, and search the offered items. The list includes notebook cells and available workspace items such as datasources, catalog tables, knowledge entries, skills and files. Select the item from the list; names can include prefixes such as catalog_table: or skill:.

For a follow-up about an existing calculation or chart, use that cell's Tag in chat action. In View, Ask about this opens Ask with the cell tagged. This is more useful than referring to "the total" when the notebook has several.

A mention points to an item; it does not prove that a rule was loaded or applied. State a critical business rule in your question, and check the resulting work. See Context and accuracy for how context is used.

3. Ask a concrete question

Start with the metric and comparison or period you need. For example:

What was paid revenue last month, broken down by product?

If your team has a specific definition, add the part that could change the answer: "Use payment date in UTC and exclude refunded orders." You do not need to prescribe the whole SQL query. If the definition is not yet shared context, add it for reuse.

Press Enter or use Send, then wait for the response. Ask to see a query or calculation when you need to check how the answer was produced.

4. Inspect the answer and refine it

Open Notebook. Use Expand code on a collapsed calculation cell and check the source alongside its output:

• Did it use the intended datasource, tables and metric definition?
• Are the dates, filters, timezone and units right?
• Did the relevant cells run successfully, and does the displayed result belong to that execution?

A confident message is not proof that a calculation is correct. Older outputs can remain visible after source edits until cells run again. Compare an important result with a trusted report or query; Verify an answer walks through that check.

If the answer needs work, name the difference and ask for a specific change in the same editing conversation. For example:

The revenue total includes refunded orders. Exclude them,
rerun the calculation, and show the revised total by product.

Tag the relevant cell if there is more than one result. Inspect the revised code, execution and output. Keep related follow-ups in this Explore. Start New Explore for an independent question or different datasource selection. Editing follow-ups can change the saved notebook; Edit a notebook covers direct cell changes.

Practice with a known answer

This small exercise lets you check both a first result and a refinement without a datasource query. Send:

Use only these sample rows; do not query a datasource.
Each row is one order. Amounts are USD and dates are UTC.
order_id,paid_at,status,amount
A,2026-09-01,paid,20
B,2026-09-02,paid,30
C,2026-09-03,refunded,5

Create a visible notebook calculation for total paid revenue from
2026-09-01 inclusive to 2026-10-01 exclusive. Include only status = paid.
Show the included order IDs, the calculation, and the total in USD.

The expected check is orders A and B, totaling 50 USD. Check the notebook rather than relying on the agent's message. Then tag the calculation cell and send:

Keep the same sample rows and paid-only rule. Narrow the period to
2026-09-01 inclusive to 2026-09-02 exclusive, in UTC.
Update the existing calculation and rerun it. Show included order IDs
and the new total in USD.

Now only A belongs in the period, for 20 USD. These totals are checks for the supplied rows, not promised model responses.

Follow up from a board

In View, select Ask, or Ask about this on a result. Ask about the tagged cell and wait for the reply before sending another question. View can inspect and run existing cells, but cannot author new source cells or change the source Explore's title.

View chat belongs to that live session. It is separate from the saved editing conversation and is not saved back to the source Explore. To keep the discussion as editable work, tag a cell and use Continue in a new explore. Follow Explore from a board for the steps and copy limits. See Explores for the Edit and View contract.

Follow up from Slack

If your workspace uses the Querio Slack app, check Settings > Slack. If Connect Slack appears, select it and complete Slack authorization. A datasource is needed before the connection can finish. Under How Querio answers in Slack, check the Datasources and Model used for new Slack Explores. Your new Explore selection in the app does not set these defaults, and changing Slack defaults does not reset an existing thread's Explore.

Make the app available in the channel where you want to ask.

1. Mention @Querio in the channel with a question such as @Querio show paid revenue last month by product, excluding refunds.
2. Wait for the reply. Mention @Querio again in the same Slack thread to refine that analysis. A new thread starts a separate Explore.
3. Select View explore in Querio in the reply to inspect the notebook, query and output. Verify the answer before using it.

If the app is unavailable or uses the wrong data, ask the person managing the integration to check Settings > Slack, the channel and the selected datasources. A Slack reply does not establish that every viewer can query the same data in Querio.

If the result or controls are unexpected

After a correction or context change, check the analysis again before sharing it.