Turn On dbt's Built In AI for One Click Docs, Tests, and SQL

1. If you don't already know your account's plan, check with whoever manages your dbt Cloud billing (a team lead or your platform team).
2. If you're on the free Developer plan, Copilot isn't available. You'd need your organization to upgrade, which is a budget conversation above most individual engineers' pay grade, so raise it with your manager rather than trying to enable it yourself.

1. An admin on your dbt Cloud account needs to opt in to Copilot at the account level before it appears for anyone.
2. If you're not an admin, ask whoever manages your dbt Cloud org to turn it on. This is a one-time setup step for the whole team, not something each engineer does individually.

Troubleshooting: If you don't see any Copilot option after this step, you may not have refreshed your session. Log out and back in to the dbt Cloud IDE.

1. Open the Studio or Cloud IDE and navigate to a model's SQL file.
2. Look in the Console section, below the File Editor, for a "dbt Copilot" option.
3. Click it to see the available AI actions.

What you should see: A set of options including Generate Documentation, Generate Tests, Generate Semantic Model, and Generate Metrics, scoped to the model you have open.

1. With a model open, click "Generate Documentation" in the Copilot panel.
2. Review the drafted description and column level descriptions it proposes for your schema.yml.
3. Edit anything that's wrong or too generic before accepting. Copilot is reading your SQL and existing metadata, not your intent, so it can miss business context you'd add by hand.

What you should see: A schema.yml block with a model description and per-column descriptions, ready to drop in or edit inline.

1. Click "Generate Tests" from the same Copilot panel.
2. Review the proposed dbt tests (not_null, unique, accepted_values, relationships) against what you know about the data.
3. Accept the ones that match your business rules, and add any edge case Copilot missed. It works from the model's structure, so a rule that lives only in a stakeholder's head, not in the SQL, won't show up automatically.

What you should see: Draft test YAML you can insert directly into your schema.yml, generated in seconds instead of the 30-45 minutes this normally takes by hand.

1. In the SQL file, use the Copilot option to generate or rewrite SQL from a natural language description.
2. Describe the transformation you want in plain language rather than writing the SQL yourself first.
3. Review the generated SQL against your project's actual column names and grain before running it.

What you should see: SQL that references your real table and column names, since Copilot has access to your project's lineage and metadata, not a generic sample schema.

1. dbt Copilot reads your project's metadata, SQL, and lineage, not row-level warehouse data, according to dbt Labs' own documentation.
2. Even so, confirm with your data governance or security team before turning this on for a production project with sensitive schemas, since AI tooling policies vary by company and this is worth a quick check rather than an assumption.
3. Never type real customer values, account numbers, or other sensitive content into the SQL generation prompt, even as an example. Describe the shape of the data instead.