Ramp Launches Glass, Positioned as AI Workspace for All Employees with Early Access
Ramp launched Glass on October 13, positioning it as an AI workspace for all employees and opened early access.
Co-founder and Co-CEO Eric Glyman responded to the question of "how Ramp moves so fast" in the announcement, stating the answer is "We learn things once." Glass is pre-configured with Slack, Notion, and internal tools via SSO on the first day of employee onboarding, eliminating the need for individual setup.
The platform includes a marketplace of over 350 reusable skills, and workflows can be saved for company-wide reuse once validated. Conversations retain memory, and automated tasks can push results directly to Slack. 323 employees (covering product, customer experience, sales, marketing, and analytics roles) have completed over 6,500 edits to the Glass knowledge base.
Feature request handling has increased from about 300 manual entries per month in Salesforce to 20,000 per month, with prompts saved by account managers automatically extracted and prioritized from customer calls. In one week in May, Ramp spent $1.5 million on tokens, subsequently migrating internal work to the self-developed Ramp Router, resulting in a 30% overnight drop in billing.
The company reports that 99.5% of employees have adopted Glass. Previously, Ramp launched AI Token Spend Management on July 16 to unify management of token spending across Anthropic, OpenAI, Gemini, and Cursor, setting limits and alerts; on August 19, it launched Router to route requests to the lowest-cost models meeting quality thresholds, which has handled Ramp's own AI traffic for three years.
Buyers are corporate finance and operations teams seeking to convert individual AI usage into organizational knowledge compounding; the seller is Ramp, productizing its internal tools. The flow of funds extends from pure expense management to AI inference routing and workspace layers, benefiting companies needing to reduce token costs and accelerate internal processes, while putting pressure on vendors that only provide single-point AI interfaces or lack organizational memory.
Ramp's recent valuation has risen from $44 billion to the $60 billion range, expanding from corporate cards to billing, procurement, and financial automation platforms, with Glass further extending its competitive scope into general AI workflow tools.
Source: Public Information
ABAB AI Insight
Ramp was founded by Eric Glyman and Karim Atiyeh in 2019, initially entering through corporate credit cards with the core proposition of helping companies reduce spending rather than increase consumption. It has since expanded into billing payments, procurement, travel, and financial automation, building the Inspect system internally to handle over two-thirds of code development. By 2026, its valuation is expected to rise from $44 billion to $60 billion in a matter of months, indicating a continued concentration of capital towards AI-capable financial infrastructure.
In terms of capital flow, Ramp first productized its internal token spending monitoring and routing tools (AI Token Spend Management and Router), and then opened the company-wide AI workspace Glass to external access. The motivation stems from its token spending reaching $1.5 million in a single week in May, achieving a 30% cost reduction through Router, and observing that individual AI usage could not automatically translate into organizational knowledge. Resources are shifting from internal efficiency tools to a sellable AI harness layer, directly interfacing with existing corporate model expenditures.
This is analogous to Stripe's continuous extension from payment infrastructure to developer tools and routing layers, or Snowflake entering broader workloads after productizing internal data capabilities. Currently, Ramp is in an expansion phase from financial control to general knowledge workflow control, with competitors shifting from traditional fintech to AI platforms with organizational memory and skill reuse.
Essentially, this represents a restructuring of the industry chain: AI costs are transformed from invisible individual token consumption to monitorable, routable, and reusable organizational-level workflows. The mechanism is that when individual employees' effective prompts and workflows can be saved and inherited by all, the marginal cost of knowledge production decreases, while general chat interfaces, lacking persistent memory and skill markets, cannot achieve the same compounding. Pricing power shifts from model providers to platforms that master corporate context and workflow standards.
ABAB News · Cognitive Laws
Individual acceleration does not equal organizational intelligence
Tools must be used internally before being sold for compounding to occur
Memory and skill sharing = organizational leverage