Y Combinator54 min
How To Build The Future: Eric Glyman, Co-CEO of Ramp
Fireside Chat, Interview
Ramp's Evolution and Strategic Pivot
- Ramp transformed from a corporate card startup founded seven years ago into a $60 billion platform managing financial operations for tens of thousands of businesses.
- The company operates at a pace described as "remarkable" by leveraging AI agents to automate financial workflows that previously required significant human intervention.
- Eric Gleiman, Ramp's CEO, notes that the company is currently in "Chapter Two," shifting focus from merely saving costs to accelerating revenue growth through AI-driven productivity.
- Ramp's median customer now saves 7.5% annually on expenses, a significant increase from the 5% savings recorded a year prior.
- The company recently achieved over $2 billion in annual revenue while growing more than 25% faster than the previous year, defying the typical gravity that slows large enterprises.
Origins of Paribus and Ramp
- Paribus, founded by Gleiman and Kareem in YC Summer 15, was an early "prehistoric AI agent" that scanned emails for price drops to automatically secure refunds.
- Paribus achieved nearly one million daily users within a year of graduation and was subsequently acquired by Capital One.
- Working at Capital One revealed systemic misalignments in financial incentives, where institutions optimized for company profit rather than customer savings, driving the decision to build Ramp.
- The core value proposition of Ramp is to compete on "value" (saving more dollars in the customer's account) rather than "price" or points schemes.
- Ramp's initial zero-to-one insight involved moving intelligence from the observation layer (email) to the operational layer (moving money), allowing the system to prevent unwanted charges rather than just refunding them later.
Internal AI Transformation and Organizational Structure
- Ramp has adopted a strategy where the top 1% of users consume 80% of AI tokens, creating a "100x engineer" productivity tier that outpaces traditional team ratios.
- To prevent organizational friction caused by uneven AI adoption, Ramp paused hiring for a full quarter earlier this year to force a cultural shift toward AI competency.
- The company deployed "Inspect," a coding agent that now processes over 85% of pull requests, with engineering pull requests growing at 40% monthly.
- Ramp's "Glass" platform serves as a centralized "company brain," ingesting data from finance, customer support, and analytics to enable shared skills and decision-making visibility.
- Non-engineers at Ramp now generate approximately 7,000 pull requests per quarter, a volume that was functionally zero two years ago, demonstrating a shift toward democratized development.
- An "expense policy agent" now approves or declines expenses with 99% accuracy, learning from overrides and adjusting policies automatically as company rules evolve.
- Ramp has stopped selling tokens and now buys them, focusing on measuring the semantic ROI of AI spend by connecting costs directly to business outcomes like conversion rates.
AI Market Dynamics and Future Outlook
- Gleiman argues that Jevons Paradox has broken regarding intelligence; despite the cost of AI dropping, societal absorption has hit a limit due to internal politics and comfort with legacy systems.
- The "99%" of organizations refusing AI adoption are often victims of siloed information and coordination costs that are too high to manage without new tools.
- Gleiman posits that the future will likely see "hyper-specialization," where thousands of firms can scale to billions in revenue by using agents to overcome human cognitive limits.
- Ramp predicts that companies must evolve from being a "system of record" to becoming an "AI harness" or they will face obsolescence.
- Ramp is currently onboarding external customers to use its internal "Glass" AI harness to manage their own company-wide workflows and knowledge bases.
- MCP (Model Context Protocol) usage at Ramp is growing 10% per week, indicating an explosion in agentic workflows where models interact with external tools autonomously.
Cultural Insights and Hiring Philosophy
- Gleiman advocates for "Sutton's Law" in hiring, suggesting recruiters find talent where proof of work is evident (e.g., builders of private Minecraft servers, math olympiad winners) rather than relying on resumes.
- The most effective leadership practice identified is the ability to say "I need help," which removes the stigma of vulnerability and accelerates problem-solving.
- Ramp maintains a culture of experimentation where employees can choose their own tools, provided they integrate into the company's standardized environment.
- The company has reduced reliance on bureaucratic review processes, allowing real-time feedback loops similar to Toyota's "Kaizen" manufacturing philosophy.
- Ramp's product release velocity has increased from one or two major releases a year to a major release every single month over the last five months.