Ramp Research: How Ramp Built an Agentic Data Analyst

Ramp built Ramp Research, an internal Slack-based AI data analyst that used an agentic architecture to explore data, inspect rows, construct and revise SQL queries, and answer more than 1,800 questions from 300 users in six weeks, increasing analytics question volume by an estimated 10–20 times while reducing response times from hours to minutes. Connected to dbt, Looker, and Snowflake, the system combined structured metadata with prose documentation written by domain experts because schemas alone could not convey business definitions and logic. Ramp evaluated the agent by validating intermediate reasoning steps, such as tool calls, table references, and query structure, rather than relying solely on final-answer checks, and included CSV previews to support user validation. The case study emphasizes that agentic analytics can accelerate decisions without replacing analysts, but it also identifies ongoing context maintenance, data quality, human oversight, and robust evaluation as operational requirements. It notes that Ramp excluded PII by design but did not publicly explain its access-control, credential-management, audit-logging, or enforcement mechanisms, highlighting governance gaps that enterprises must address through scoped identities, permissions, monitoring, and managed infrastructure such as MCP and agent gateways.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.