Skip to main content
MindStudio
Pricing
BlogAbout
My Workspace
Ramp CEO Eric GlymanRamp AI agentsRamp valuation

How Ramp's Eric Glyman Went From Email Hacks to a $60B AI Fintech

Ramp CEO Eric Glyman on building a $60B fintech by aligning incentives with customers, from Paribus's email agents to AI-driven finance tools.

Edited by Luis Chavez-Mattos, Director of Product RSS
How Ramp's Eric Glyman Went From Email Hacks to a $60B AI Fintech

What is Ramp and why does its origin story matter?

Ramp is a corporate card and spend management platform now valued at $60 billion, built on a simple premise: save companies time and money on their finance operations. That premise did not come from nowhere. Ramp’s co-founders, Eric Glyman and Karim Atiyeh, spent their first startup, Paribus, building what Glyman now describes as an early AI agent, one that scanned email receipts and automatically asked for refunds when prices dropped. The lessons from that company, and from the two and a half years the founders spent inside Capital One after selling Paribus to it, directly shaped how Ramp makes money and what it automates today.

TL;DR

  • Ramp’s founders learned the agent playbook at Paribus, a savings app that scanned users’ Gmail and Yahoo inboxes for purchase receipts, checked prices automatically, and emailed retailers for refunds on the customer’s behalf.
  • Capital One acquired Paribus, giving Glyman and Atiyeh roughly two and a half years inside a major financial institution, where they saw how incentive structures around credit card points and rewards could work against the customers those products were supposed to serve.
  • Ramp’s founding insight was a business model choice, not just a technical one: build financial products designed to help customers spend less, rather than products that profit from customers spending or forgetting to cancel subscriptions.
  • The company’s first wedge was a corporate card that doubles as expense software, aimed at surfacing wasted spend, like unused subscriptions, and cutting the manual work of expense reports.
  • Ramp now runs finance operations for tens of thousands of businesses and is moving deeper into automating that work with AI agents rather than just flagging problems for humans to fix.
  • The founders’ earlier Reddit and referral-driven growth tactics at Paribus, including a two-sided referral discount that cut both the company’s and the referrer’s fee, show an early pattern of aligning incentives with users that carried into Ramp’s design.
  • A recurring theme in Glyman’s account is the gap between seeing a problem and being able to act on it, which he says pushed Ramp toward embedding automation directly into the movement of money rather than just alerting customers to issues.
Build enterprise applications with AI that are
Secure.Role: Adjuster✕ Export PHI
Regulation-compliant.✓ HIPAA✓ SOC 2
Auditable.09:14M. Chen approved claim #4471
Scalable.12,480 users online

How did Paribus work, and what did it teach the Ramp founders?

Paribus started from a personal frustration. Glyman noticed that a flight he and friends booked had dropped in price the next day, and after reading the airline’s policy, he realized a refund was available just by asking within 24 hours. That single $1,000 refund for a group of friends became the seed of a company: automate that process across online retail, flights, and hotels.

The mechanics were straightforward in concept. Paribus connected to a user’s email, scanned inboxes for purchase receipts, structured that data, checked whether prices had dropped, and if so, sent a refund request from the user’s own account. Glyman calls it one of the “prehistoric AI agents,” built initially with scripts in VBA, long before the current wave of LLM-based automation.

Growth came from scrappy, unglamorous channels. The founders carried signs around New York, exhausted their friend networks fast, and found real traction on Reddit’s “Beer Money” subreddit, where a single post describing the app brought in the first wave of users. From there, user interviews (including posing as outside market researchers to hear how real users described the product to friends) revealed a trust problem: people loved a tool that saved them money but got wary when they learned it kept 25% of the savings. The fix was a two-sided referral structure that lowered the cut for both the referrer and the new user, which accelerated growth through word of mouth.

Why did the founders sell to Capital One, and what did that teach them about finance?

Within about a year of going through Y Combinator (Glyman notes they applied twice before getting in), Paribus had grown to nearly a million daily users and drew an acquisition offer from Capital One. The founders took it and spent roughly two and a half years inside the company.

That stint became as formative as the startup itself. Glyman describes watching a “world-class organization” operate at scale, including both strong processes and the kind of institutional inertia that can smother good ideas. The more pointed lesson was about incentives: credit card businesses often profit from rewards programs (points, cash back) in ways that depend on customers either overspending to earn rewards or under-using the rewards they’ve earned. Different teams optimized different metrics, like purchase volume or net margin on rewards costs, without a shared incentive to actually make the customer’s financial position better.

User research conducted while still inside that system surfaced a simple but pointed insight: when asked what would make them use a card more, customers didn’t necessarily ask for a better points multiplier. They said they wanted more money in their bank account, period. That reframing, that an app or card could compete on raw dollar value instead of gimmicks, became the founding thesis for Ramp.

What is Ramp’s actual business model?

Developers have their tools.Enterprises have Remy.
Developers> add rate limiting to the API
● Reading src/server.ts
● Editing 3 files
+ import { limit } from 'rate'
- app.use(cors())
✓ Tests passed · ready for PR
Enterprises
Loan Approvals● Live
SOC 2Role-based accessSelf-hosted
Mortgage · $412,000Approved
Small business · $85,000Approved
Auto · $31,500Approved

Ramp’s answer to that incentive problem was to design a company around customer savings instead of around maximizing the profit pulled from each transaction. Concretely, that meant building a corporate card that doesn’t just process payments but actively finds wasted spend, flags subscriptions nobody uses, and automates expense reporting so finance teams spend less manual time reconciling receipts.

Glyman frames this as a business model decision before it was a technology decision: instead of running thin margins on card interchange or rewards, Ramp routes the profit pools generated by its financial products back into building software, like bill payment tools, that further reduces customer cost and effort. The pitch to customers is not a rebate percentage; it’s fewer dollars spent and fewer hours spent managing spend.

How is Ramp using AI agents in finance operations?

Ramp has grown into a platform that manages financial operations for tens of thousands of businesses, and the company’s current direction leans heavily on agents performing that work rather than only surfacing it for a human to review. The thread connects directly back to Paribus’s limitations: that earlier company could see which purchases were eligible for refunds across hundreds of millions of scanned emails, but its only lever to act was sending a customer-service email and hoping for the best.

Ramp’s position, closer to the actual movement of money through a corporate card and expense platform, removes that bottleneck. Instead of just detecting an unused subscription or a reconcilable expense, the system is built to act on it: canceling, categorizing, reconciling, or flagging for payment without someone manually pushing every step through. That shift, from detection to autonomous action, is the throughline Glyman draws from a decade-old email-scanning tool to a $60 billion finance platform.

Is the “100x employee” framing realistic?

The idea behind building toward a dramatically more productive finance function is that software doing the repetitive, structurable parts of financial operations (categorization, policy checks, refund requests, reconciliation) can free human finance teams to spend time on judgment calls instead of data entry. Whether any given team gets an actual 100x productivity multiplier depends heavily on how much of their workload is structurable in the way Paribus’s refund logic or Ramp’s expense rules are. The more a finance process resembles “check a condition, take a defined action,” the more agent automation can realistically absorb it. Processes requiring negotiation, nuanced judgment, or exceptions still need people in the loop.

Frequently Asked Questions

Who founded Ramp and what did they do before it?

Ramp was co-founded by Eric Glyman and Karim Atiyeh. Before Ramp, they built Paribus, a savings app that scanned users’ email for purchase receipts and automatically requested refunds when prices dropped. Capital One acquired Paribus, and the founders worked there for roughly two and a half years before starting Ramp.

What problem is Ramp trying to solve?

Ramp is built around the idea that financial products, like corporate cards, are often designed to profit from customer spending rather than to reduce it. Ramp’s model ties its product design to cutting customer costs and time spent on finance operations, using that alignment as its core value proposition.

How did Paribus use AI-like automation before the current LLM era?

› Build a claims-review app for 400 adjusters.
already included:✓ HIPAA✓ access control✓ audit log✓ self-hosted
The enterprise platform for building software with AI.

Paribus connected to users’ email accounts, scanned for purchase confirmations, checked whether prices had since dropped, and automatically sent refund requests from the user’s account. Glyman describes it as an early, “prehistoric” form of an AI agent, built with simple scripts rather than modern language models.

What is Ramp’s current valuation?

Ramp is valued at $60 billion, according to Glyman, built up from its original corporate card product into a platform used by tens of thousands of businesses for finance operations.

How does Ramp differ from traditional credit card companies?

Glyman argues traditional card companies often profit from reward schemes that benefit from customers overspending or under-redeeming points, with teams across the business optimizing for different, sometimes misaligned, metrics. Ramp’s model instead ties profitability to the software it builds for reducing customer spend and manual finance work, rather than to rewards mechanics.

Editorial standards

Presented by MindStudio

No spam. Unsubscribe anytime.