Keep dbt as your source of truth. Your definitions guide every answer in Claude or ChatGPT, queries only touch the tables you approve, and each team sees only what it should.
“Why did paid search drop?” goes into their queue, behind the last ten.
Your dashboards answer the first question. Nobody can ask the next one.
People paste raw tables into ChatGPT, and your data team cleans up the wrong numbers.
Models and semantic models
Converted to open OSI
Before anything changes
In Claude, ChatGPT and agents
What changes
One dbt project. Your team asking its own questions.
Start with the questions your data team answers every week and the reports someone builds by hand. Then open it up team by team, on the same definitions.
No. The CI step runs dbt parse, which is part of dbt Core.
AI drafts them from the tables your dbt project builds, and your team reviews them before anyone relies on them.
Snowflake, Redshift, Postgres and MySQL today. On BigQuery or Databricks? Tell us on the call and we’ll check your setup.
They work together. dbt defines your models. FrontierOps puts them next to your other systems, with access per team, an audit trail, reports and agents, in any AI client.