How Ramp's internal AI agents crossed a million sessions

Ramp data platform engineer Jay Sobel described the company’s “internal AI” strategy at dbt Summit 2026 as an effort to equip employees across the organization with productive large language model tools, a role he views as a natural extension of analytics engineering. Ramp has developed several internal platforms, including Ramp Research for data questions, Inspect for sandboxed coding-agent sessions that has exceeded one million uses, Glass for employee access to internal systems, Rampplify for rapidly deploying internal apps, and WaveDash, a Git-based SQL and Markdown dashboarding system intended to replace Looker. Data-agent query volume has grown from roughly 10,000–15,000 monthly questions to levels exceeding comparable human-driven querying, while adoption has expanded beyond engineers to teams such as sales. Sobel emphasized avoiding dependence on any one AI provider, acknowledged that the company does not yet rigorously measure AI return on investment, credited security engineering for enabling the effort, and advised smaller teams to rely on existing design systems, chart libraries, and simple internal repositories rather than building comprehensive custom platforms.

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.