Answer company data questions in Slack with an analyst agent grounded in dbt docs

Ramp built Ramp Research, an internal analyst agent that answers data questions in Slack within minutes instead of leaving them queued for an on call analyst. It draws on indexed metadata from dbt, Looker and Snowflake plus documentation written by each domain owner, and it can inspect column values, branch and backtrack across thousands of tables rather than relying only on vector search. Threads keep state for follow up questions and include CSV previews, and the agent has no access to personal data. A Python test framework in their dbt project checks answers and the intermediate tool calls. Since its August launch it has answered over 1,800 questions for 300 users, and in four weeks it handled 1,476 questions against 66 in the old help channel.