Platform · dbt
Transformations you can test, review and trust.
dbt brings software engineering discipline to the transformation layer: models in version control, tests that run on every change, and documentation and lineage generated from the code itself, all pushed down to Snowflake.
Different ways of working, not a longer feature list.
Every transformation is code in a repository. Changes are reviewed, history is kept, and rolling back is a git operation rather than an archaeology project.
Uniqueness, referential integrity and business-rule assertions run on every change. Broken data stops in CI, not in a meeting.
The model is the documentation. Column descriptions, dependencies and a full lineage graph come out of the same source that runs in production.
A metric or a transformation lives in one model. Every downstream tool, including Sigma and Power BI, reads the same definition.
Faster to the first correct model, from three angles.
dbt Copilot isn't the only AI in the loop. Claude Code and Snowflake's own Cortex Code can scaffold and review dbt work too, each suited to a different part of the workflow, all landing as reviewed, version-controlled code.
dbt Copilot
AI native to the dbt project
Describe the model you need and Copilot proposes the SQL, the tests and the documentation. The output lands as code in the project, so it is reviewed and versioned like anything a person wrote.
- Generates models, tests and docs from a prompt
- Works inside the version-controlled project
- Output is reviewed like any other change
- Speeds the build without skipping the discipline
Claude Code & the Agent SDK
Agentic coding across the whole project
Claude Code works across the full dbt project from your terminal or IDE: refactoring models, fixing failing tests, writing migrations and reasoning over the codebase and lineage graph, not just drafting a single new model.
- Operates across the whole repo, not one file at a time
- Refactors and fixes, not just drafts
- Same model that powers Cortex Agents and CoWork
- MCP connects it to your governed Snowflake data
Cortex Code
Snowflake's own agentic coding assistant
Cortex Code runs inside Snowflake's perimeter and understands your warehouse, so it can scaffold and modify dbt models against your real, governed schema without the project or the data ever leaving Snowflake.
- Runs inside Snowflake: no code or data leaves the perimeter
- Understands your live schema and lineage, not just the repo
- Fits where governance rules out an external IDE agent
- Complements Claude Code rather than replacing it
Not the syntax. The reasons analytics engineering on dbt pays back.
Transformation
The data team starts working like a software team: reviews, tests, releases, and the transformation layer stops being the place where trust quietly breaks.
Business process improvement
Business logic is defined once, tested continuously and documented automatically. The definition of “active customer” stops being an argument.
Optimisation
Transformations are push-down: Snowflake does the compute, dbt orchestrates. Incremental models mean you reprocess what changed, not the whole history.
Cost saving
Less rework from broken data, less time spent reverse-engineering undocumented SQL, and no separate transformation engine to license and run.
Future-proofing
A tested, documented, version-controlled model layer is the foundation every downstream tool and AI agent depends on. It is the asset that outlasts the dashboard.
A way of work, not a one-off delivery.
We care as much about the how and the why as the what. Setting up dbt properly is how we invest in your team's development practices, so you can keep extending the platform long after this project ends, backed by real training and support, not a handover document.
Production-grade from day one
Project structure and standards set up the way a mature data team would run them, not left to grow organically into technical debt.
CI/CD built in
Environment separation and CI/CD configured from the start, so a change is a reviewed pull request, not a risk taken in production.
Tested and documented
Testing and documentation are part of the workflow, not an afterthought bolted on once something has already broken.
Core to the platform
Think, Build, Grow: the model layer is core to how we build, not an add-on we reach for when asked.
Training and mentorship built in
Structured dbt training plus senior engineers pairing with your team sprint by sprint, so the way of work sticks after we leave, not just the code.
