Key Moments

Ramp’s Bet on the 100x Employee

Y CombinatorY Combinator
Science & Technology7 min read55 min video
Nov 18, 2026|17,045 views|183|6
Save to Pod
TL;DR

Ramp's AI agents are revolutionizing financial operations, but their internal AI tools are primarily benefiting the top 1% of users, raising questions about broader adoption.

Key Insights

1

Ramp's platform now manages financial operations for tens of thousands of businesses, with AI agents playing an increasing role.

2

Paribus, Ramp's predecessor, automatically recovered money when online purchases went on sale, and saw a 10% weekly growth rate by offering a two-sided referral deal.

3

Ramp's co-founder, Karim, believed that a simple value proposition and a focus on helping individuals spend less, rather than just making more money for the company, could lead to a different kind of business.

4

The average company using Ramp saves 7.5% annually on its expenses, which is several times more than interchange fees.

5

Only 1% of users are consuming 80% of AI token spend, and within Ramp, engineers using AI tools saw productivity gains of 100x, while designers saw 2x, and product team members saw 5x.

6

Ramp's internal AI tool, Glass, is now used by 1925 out of 1930 employees and automatically connects to all company systems, providing a unified view of data and enabling faster decision-making.

From Paribus to Ramp: A Vision for Value-Driven Finance

The journey of Ramp's co-founder, Eric Gleiman, began with Paribus, an early AI agent that automatically recovered money for users when online purchases dropped in price. After a successful stint with Y Combinator and an acquisition by Capital One, Gleiman and co-founder Karim developed a core insight: financial products should be designed to help individuals spend less, not just generate more revenue for the company. This philosophy, centered on competing on value rather than price, laid the groundwork for Ramp. The initial idea for Paribus stemmed from a personal frustration with fluctuating flight prices, leading to the realization that automated price monitoring and refund requests were feasible. Early user acquisition for Paribus involved unconventional methods, including posting on Reddit's 'Beer Money' subreddit, and a key learning was the power of aligning incentives through a two-sided referral program, which dramatically boosted growth.

Rethinking Financial Models: Competing on Value

Gleiman and Karim observed that traditional financial institutions, including credit card companies, were often structured to maximize company profits, sometimes at the expense of customer value. This was exemplified by reward programs that encouraged spending but made points difficult to redeem or devalued them. They identified a significant opportunity to create a '100x better' value proposition by focusing on directly increasing customers' available cash, rather than offering fractional rewards. This meant fundamentally rethinking the business model. Instead of taking a small margin from transactions, Ramp aimed to leverage the efficiencies gained from helping customers spend less to invest in superior technology and software. This approach allowed them to offer better core products, such as a card that not only reduced spending but also managed expense reporting, and subsequently, bill payments, freeing up more capital for investment compared to competitors focused on incremental gains.

The 'Zero to One' of Ramp: Simplifying Financial Chaos

The foundational 'zero to one' insight for Ramp, similar to Paribus, was identifying and simplifying areas of financial waste for businesses. This included the often-overlooked problem of managing numerous subscriptions and unused services. However, Ramp moved beyond just visibility to action, focusing on the ability to intervene in real-time and prevent unnecessary spending. This operational capability, acting directly on financial flows, was a critical differentiator. Unlike Paribus, which relied on post-purchase refunds, Ramp aimed to stop the money from leaving in the first place. This led to building integrated financial operations tools that consolidated disparate systems for credit cards, expenses, bills, and travel into a single, streamlined platform. This integration not only reduced administrative overhead but also provided a clearer financial picture, enabling companies to save significantly. The average Ramp customer now saves 7.5% annually on expenses, a substantial improvement over previous interchange fee structures.

AI as an Operational Layer: Enhancing Productivity and Growth

Ramp's vision extends to leveraging AI not just for product features but as a fundamental operational layer for businesses. The rapid advancements in AI models, from GPT-3.5 to more sophisticated versions, have underscored the need for organizations to re-evaluate their structures. Internally, Ramp has developed AI tools like 'Glass' (a knowledge and workflow system) and 'Inspect' (a software agent for code reviews). These tools have led to dramatic productivity gains: engineers have seen 100x improvements, designers 2x, and product team members 5x. However, a significant disparity exists, with the top 1% of users consuming 80% of AI token spend. This highlights a challenge in broader AI adoption, where internal policies and skill gaps may be limiting for the remaining 99% of employees. The company's strategy is to make AI integration seamless, aiming to turn every employee into a highly productive 'engineer-level' contributor.

The 'Glass' Ecosystem: Unifying Company Knowledge and Operations

Glass, Ramp's AI-powered knowledge and workflow system, serves as the company's 'brain,' connecting to all internal systems, including data warehouses, analytics platforms, and customer support tools. This integration provides a unified, comprehensive view of company operations, enabling faster and more informed decision-making. For individuals, Glass offers personalized access to information and workflows, with features like automated pull request generation by non-engineers, which has seen a dramatic surge. This democratization of advanced capabilities, like code generation and policy adherence, significantly speeds up product development and problem-solving. For instance, a customer support issue that might have taken months to resolve in a traditional setup can now be addressed by Inspect within minutes, dramatically improving efficiency and customer satisfaction.

Navigating the AI Transition: Culture, Adoption, and the Future of Work

The transition to widespread AI adoption presents significant cultural and operational challenges. Gleiman likens the difficulty of learning new AI tools to switching operating systems, emphasizing the need for time and support. He notes that within Ramp, the productivity gap between early AI adopters and others is substantial, creating organizational friction. The company's approach is to foster an experimental culture, encouraging employees to explore and utilize AI tools like Claude and Cursor, rather than imposing rigid policies. This philosophy is reflected in their hiring practices, which prioritize demonstrated ability to build and solve problems over traditional resumes. The key takeaway for organizations is that AI adoption isn't just about new software; it's a fundamental shift in how work is done, requiring a top-down commitment to learning and adaptation. The expectation is that as AI becomes more capable and cost-effective, it will enable new forms of business and unprecedented levels of productivity.

The Evolving Role of AI in Business and the 'AI Productivity Loop'

Ramp views itself not just as a fintech company but as a productivity company, focused on helping businesses spend less and achieve more. The argument that companies must either die as record-keeping entities or live long enough to become AI tools is central to their strategy. Tools like Glass are designed to harness AI's power, enabling employees to perform tasks previously requiring specialized skills. The 'AI productivity loop' is described by Gleiman: as AI becomes cheaper and more capable, its adoption increases, leading to further innovation and cost reductions, creating a virtuous cycle. He highlights that the past two months have shown a breakdown of Jevons Paradox, where increased efficiency and lower costs haven't necessarily led to a proportional increase in consumption of AI, suggesting that broader organizational capacity for adoption is the current bottleneck. This is a critical challenge that companies like Ramp aim to address.

Hiring for Builders and Problem-Solvers in the AI Era

Ramp's hiring philosophy, exemplified by 'Sutton's Law,' emphasizes finding individuals with a proven track record of building and problem-solving, rather than relying solely on traditional credentials. This involves looking for evidence of initiative and capability, such as teenagers who have successfully built businesses on platforms like Minecraft or won competitive math and informatics olympiads. The focus is on tangible achievements that demonstrate a candidate's passion and skill in creating value that others want. This approach aims to identify individuals who are inherently driven to build and innovate, which is crucial in a rapidly evolving technological landscape. The ultimate goal is to foster a culture where seeking help and collaboration is encouraged, enabling individuals and the company to overcome obstacles and achieve significant breakthroughs.

Common Questions

Eric Gimmon's first startup was Paribus, an AI agent that automatically recovered money for users when online purchase prices dropped. It worked by scanning email receipts and initiating refund requests.

Topics

Mentioned in this video

Companies
Ramp

A financial operations platform for businesses, initially launched as a corporate credit card, now valued at $60 billion and increasingly using AI agents.

Anthropic

Mentioned as the company behind Claude Code, for which Ramp was an early beta customer.

OpenAI

Mentioned as a provider of AI models, with users having the option to choose between OpenAI models, Claude, or open-source alternatives within Ramp's ecosystem.

Y Combinator

A startup accelerator program that Eric Gimmon and his co-founder Karim applied to twice before being accepted, which significantly altered their company's trajectory.

Schwab

Mentioned in the context of long-standing issues with software and bug fixes, implying that companies like Ramp are fundamentally changing the speed and efficiency of product development.

Capital One

Acquired Paribus, providing Eric Gimmon and Karim with the opportunity to observe a world-class organization and learn about financial institutions.

Reddit

A platform where Paribus gained many of its early users through a subreddit called 'Beer Money', indicating its effectiveness in reaching specific online communities.

Snowflake

A company whose platform Ramp integrated with, demonstrating the broad range of integrations available within Ramp's ecosystem.

Cognition

A company that Ramp served when it was small, demonstrating Ramp's growth alongside exceptional companies and its early adoption of their tools.

Databricks

A data analytics platform for which Ramp was an early customer, showcasing Ramp's position at the forefront of adopting new technological tools.

Google

A company whose platform Ramp integrated with, showcasing the extensive connectivity of Ramp's systems to various essential business tools.

American Express

A company sometimes compared to Ramp, highlighting its evolution from a logistics service in 1850 to a global financial conglomerate.

Stripe

A payment processing company that Ramp integrated with, illustrating the comprehensive integration capabilities of Ramp's platform.

Microsoft

The company where Eric Gimmon began his career, serving as an example of large, established organizations with 'fiefdoms' and information silos.

Zendesk

A customer service platform that Ramp integrates with, allowing Glass to pull customer support data for a comprehensive company view.

Software & Apps
Git

A version control system mentioned in the context of allowing non-engineers to create pull requests, facilitated by tools like Glass.

Paribus

An early AI agent that automatically recovered money when online purchases dropped in price, operating by scanning emails for receipts and initiating refund requests.

Claude Code

An AI coding tool developed by Anthropic, for which Ramp was an early beta customer, highlighting Ramp's proactive adoption of new technologies.

Cursor

An AI-powered code editor that Ramp was an early adopter of, reflecting the company's strategy of experimenting with and leveraging cutting-edge tools.

Visual Basic for Applications

A programming language used in the early stages of building Paribus's refund-generating triggers.

Slack

A communication platform that Ramp integrated with, allowing for seamless workflow and data sharing across different business functions.

Claude

Mentioned as an AI model that some individuals are using for tasks like tax preparation, indicating its advanced capabilities beyond simple assistance.

ChatGPT

The AI model launched less than four years prior, initially unable to perform basic calculations but has since evolved significantly.

GPT-3.5

An earlier version of ChatGPT, highlighted for its limitations in performing mathematical operations.

GPT-4

A later version of the GPT series, noted for surpassing average human intelligence and demonstrating significant advancements in problem-solving.

Inspect

An AI software agent developed by Ramp for code reviews, capable of analyzing pull requests and identifying potential errors.

More from Y Combinator

View all 645 summaries

Ask anything from this episode.

Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.

Get Started Free