Ask AI a question about your data

Outcome: complete this walkthrough and you will have asked a natural-language question and got a live SQL-generated answer with a result table you can export · Time: ~2 min active · ~3 min elapsed (waits included)

Before you start

• Role: viewer or higher

• You have a demo-tenant account (viewer role is enough to ask)

• Your catalog has tables (the demo has ecommerce.public.orders)
• A little credit balance — Ask AI meters LLM tokens + DuckDB seconds (tiny per question)

Steps

1. Open Ask AI

Click Ask AI in the top bar (star icon).

Locator: Ask AI (top bar)

• You should see: The Ask AI panel slides in: a MODEL selector (deepseek.v3.2), suggested questions, and an input box.

• If not: Panel won't open → your account may be out of credits (Ask AI is metered).

2. Ask a question

Click the suggestion Count orders by status — it fills the input — then press the send (↑) button.

Locator: suggestion chip + ↑ send

• You should see: The question appears in the conversation and Ask AI shows Thinking… while it generates and runs SQL.

• If not: If the ↑ is disabled, type the question manually and press Enter.

3. Read the answer

Wait for the result card.

Locator: result card (Found N result(s))

• You should see: Found 4 result(s) with a table: completed 7 · pending 2 · cancelled 1 · shipped 2. A CSV download and Read aloud button are available.

• If not: An error appears → if it mentions credits, top up; if it's a product error, note the exact text and page the platform owner.

4. See the SQL it wrote

Click the </> button on the result card to reveal the generated SQL.

Locator: </> button

• You should see: SELECT status, COUNT(*) AS order_count FROM ecommerce.public.orders GROUP BY status LIMIT 50; — you can copy it or export the results as CSV.

Expected result

When the steps are done:

• model: deepseek.v3.2
• question: Count orders by status
• rows: 4
• result: completed 7 · pending 2 · cancelled 1 · shipped 2
• generatedSql: SELECT status, COUNT(*) AS order_count FROM ecommerce.public.orders GROUP BY status LIMIT 50;
• measuredOn: 2026-09-03

Start over & leftovers

Nothing is written to storage — Ask AI only reads your catalog and returns results. Your question appears in the Ask AI conversation history for your user (dismissible by clearing/starting a new chat).

Try variations

1. Try the other suggestions: 'Show me top customers by lifetime value', 'Find customers with no orders', 'Show monthly revenue trend'.
2. Ask your own question in plain language (e.g. 'how many products cost more than 20?').
3. Use CSV to download the result as a file.
4. Choose a different model from the MODEL selector (per-tenant override).

Terms used

• Ask AI
• MODEL
• deepseek.v3.2
• catalog
• generated SQL
• CSV export

_{Walkthrough W9-ask-ai · verified against build server v130 + frontend (2026-09-03) · role viewer. Scenario Pack source: walkthroughs/W9-ask-ai/. This page is rendered from scenario.json — edit that, not this.}

_{Results come from the tenant's own demo catalog tables.}