Ramp launches Glass to capture and reuse internal AI workflows

Ramp has launched Glass, an internal AI platform designed to prevent the loss of institutional knowledge by embedding successful workflows directly into daily operations. The tool addresses the inefficiency of isolated AI usage, where effective prompts and processes disappear when employees leave or close their sessions.

Capturing Institutional Knowledge

Glass allows employees to save and share their best AI interactions, turning individual discoveries into company-wide assets. For example, a prompt that automatically extracts feature requests from customer calls increased Ramp’s capture rate from 300 to 20,000 per month. Currently, 323 employees have contributed over 6,500 edits to the system’s knowledge base, covering roles from product managers to salespeople.

Cost Control and Model Routing

To manage the high costs associated with widespread AI adoption—Ramp previously burned $1.5 million in tokens in a single week—Glass integrates with Ramp Router. This component selects the most appropriate model for each task, which reduced Ramp’s AI bill by 30% overnight. The platform also features a real-time cost meter for every project, allowing teams to monitor spending as it happens rather than waiting for invoices.

Key Features

• Knowledge Integration: Connects with Linear, Notion, Google, Slack, and Gong to ground AI responses in company data.
• Multiplayer Sessions: Enables real-time collaboration by allowing teammates to join and view an ongoing AI session.
• Model Agnosticism: Supports both open and closed models, allowing users to switch to newer versions as they become available.