Department of Product: How Ramp built an internal AI productivity platform

Rich Holmes, writing on the Department of Product Substack, examined Ramp’s internally built AI productivity platform, Glass. Ramp is a corporate expense and finance management company that built Glass not as a customer-facing feature but as internal infrastructure to accelerate AI adoption across the organization.

What Glass is

Glass combines five core capabilities. The first is single sign-on and pre-configured integrations that eliminate the setup friction that typically discourages adoption. The second is Dojo, a skills marketplace where teams can build, share, and discover reusable AI workflow templates. The third is Sensei, an AI recommendation engine that suggests relevant workflows based on each user’s role and recent activity. The fourth is persistent memory that preserves context between sessions. The fifth is scheduled automations that run locally from user machines without requiring engineering involvement.

Over 350 Skills have been shared across Ramp’s organization. Sensei surfaces relevant ones to each user rather than exposing the full catalog at once, which matters: access to many tools without guidance tends to reduce adoption rather than increase it.

The central argument

Holmes frames the article around a claim from Ramp: internal productivity and speed is a moat. This reframes internal tooling from operational overhead to competitive advantage. Companies that systematically improve AI adoption across teams move faster, generate institutional knowledge about what AI does well in their context, and get cleaner signals about what to build into customer-facing products.

Building Glass required the same skills that go into building any product — understanding adoption barriers, designing for discoverability, preventing tool fatigue, and iterating on what actually changes behavior. The lessons from internal deployment become inputs to the external product roadmap.

Who it is useful for

This case is relevant for PMs and product leaders at companies actively deploying AI tools internally. The core insight is about approach rather than any specific tool: guided adoption through curated recommendations and reusable templates outperforms giving teams raw API access and expecting them to self-organize. The pattern Ramp used — skills marketplace plus intelligent surfacing plus persistent memory — is transferable to any organization trying to move from individual AI experimentation to coordinated team-level practice.