Quickstart

Use this flow when you want to get productive quickly without skipping the core review step.

1. Sign in and choose the right account

Sign in to Tukun.ai and confirm you are in the correct account before doing anything else.

Why this matters:

• data sources are account-scoped
• conversations and saved assets are account-scoped
• billing and usage are account-scoped

If you are working with production data, do not run setup in a test or personal account by accident.

2. Connect one data source first

Start with one approved source instead of trying to onboard everything at once.

Prepare:

• a read-only credential
• the approved database or schema scope
• one or two known tables you can use for validation

If you need setup details, read Connect Data.

3. Ask one concrete question

Open the Workbench and start with a question that has:

• one clear metric
• one clear time range
• one clear breakdown or filter set

Example:

Show weekly active accounts for the last 12 weeks by plan tier.

Avoid stacking too many goals into the first question. If you need more depth, get the first result right and then follow up.

If you uploaded a structured file, start with the suggested follow-up questions shown under the result. They usually give you a faster path into trend, breakdown, or anomaly checks.

4. Review before you trust

Check:

• whether the metric interpretation matches your intent
• whether the time range is correct
• whether the filters and exclusions are correct
• whether the result looks directionally plausible
• whether the evidence is strong enough for the decision you want to make

Use Review Results as the checklist.

5. Resolve ambiguity instead of sharing around it

If the answer is close but not yet trustworthy:

• rewrite the question
• add the missing filter or breakdown
• narrow the time range
• define or refine the metric in semantic modeling

Do not save a result just because it looks polished. Save it when the team agrees it is conceptually right.

6. Save only repeatable outputs

Once a result survives review:

• save it as a card if it answers one recurring question
• add it to a dashboard if it belongs in a repeated operating review
• model the metric if the same business definition keeps reappearing

A good first-session outcome

Your first successful session should end with these outcomes:

• one connected source
• one validated question
• one checked answer
• one clear list of definitions that still need cleanup

That is a better onboarding milestone than trying to create a perfect dashboard on day one.