Can an agent replace dbt?
A framework for writing, testing and documenting the SQL transformations that turn raw warehouse data into usable tables.
The framework, no. The seats, look harder. dbt Core is free and stays as the structure your transformation logic lives in. Writing the models is now agent work, and dbt Cloud is priced per developer doing exactly that.
- Indicative spend
- €300/mo
- What it actually costs
- dbt Core is free; Cloud from roughly €100/developer/mo
- Verdict
- The agent does the work inside it. The tool stays because that is where the data lives.
What the agent takes over
Every job this product exists to perform, with our verdict on each. Follow one through for the step-by-step breakdown.
- Data analysisMostly. An agent cleans, queries, models and visualises quickly and competently. The risk is not that it computes wrongly — it is that it answers the question you asked rather than the one you meant.MOSTLY
- Software developmentMostly. An agent writes, tests and ships real features in a codebase it can read, and it does so faster than a person. It cannot decide what to build, and it degrades badly as a system gets large and undocumented.MOSTLY
- DocumentationYes, and it fixes the real problem, which was never writing documentation but keeping it true. An agent regenerates docs as the code changes, which no human team has ever sustainably managed.YES
- Reporting and dashboardsYes. Pulling from several systems, reconciling the definitions, building the view and explaining what moved is exactly what agents do well — and it removes the recurring tax of someone rebuilding a spreadsheet every Monday.YES
Why it survives
dbt the standard is safe because it is free. dbt Cloud is a per-developer bill in a world where model-writing is agent labour.
What you would still need it for
- The repo of models that is your transformation logic
- Tests and lineage the agent works inside
What replaces it
- An agent writing and testing models in your dbt Core project
- A scheduler you already have for the runs
The brief
What you would tell an agent to take over from dbt, assembled from the jobs above.
I want to cut the work inside dbt. It currently does: A framework for writing, testing and documenting the SQL transformations that turn raw warehouse data into usable tables. Take over this work: - Data analysis — MOSTLY. Mostly. An agent cleans, queries, models and visualises quickly and competently. The risk is not that it computes wrongly — it is that it answers the question you asked rather than the one you meant. - Software development — MOSTLY. Mostly. An agent writes, tests and ships real features in a codebase it can read, and it does so faster than a person. It cannot decide what to build, and it degrades badly as a system gets large and undocumented. - Documentation — YES. Yes, and it fixes the real problem, which was never writing documentation but keeping it true. An agent regenerates docs as the code changes, which no human team has ever sustainably managed. - Reporting and dashboards — YES. Yes. Pulling from several systems, reconciling the definitions, building the view and explaining what moved is exactly what agents do well — and it removes the recurring tax of someone rebuilding a spreadsheet every Monday. Do not take over: - The repo of models that is your transformation logic - Tests and lineage the agent works inside These stay with me across all of it: - Framing the question - Knowing the data’s history - Deciding what to do about the answer - What to build - Architecture decisions with long consequences - Production accountability - Code review - Architectural rationale - Deciding what deserves documenting - The overall narrative - Choosing the metrics - What the numbers mean for the plan - Anything shown to investors Before we start, tell me: which of these you cannot do with the access I can actually give you, and what would break if this ran unattended for a month. — brief built at cananagentdo.com/dbt
Compare
Keep the tool, cut the hours
dbt is not the line item worth attacking. The money is in the people-hours spent working inside it, and that is what an agent takes over — with dbt still holding the data.
Put an agent on it