Ramp's internal AI stack: agents, sandboxes, and an AI-first BI tool
Ramp data platform engineer Jay Sobel breaks down the agents, sandboxes, and dashboards his team built after deciding internal AI was just analytics engineering with a bigger mandate.
Overview
Ramp's data team built interconnected AI tools including Ramp Research, Inspect, Glass, and a new dashboarding app. Covers implementation strategy, context-layer architecture, narrative-based explanations, and rollout/onboarding.
What you'll take away
- 01
Ramp's approach to AI is organizational, not product-embedded: giving every team, from engineers to sales, agents and tools grounded in the company's own data and systems.
- 02
Inspect, Ramp's sandboxed coding-agent ticket system, has passed a million sessions, and WaveDash renders every dashboard as a SQL file in a repo so agents can read and edit it the same way engineers do.
- 03
Ramp deliberately avoids locking into a single AI lab's models, citing volatile vendor valuations as a reason to keep its options open.
Session speaker
Jay Sobel
Ramp