Part IThe Story
The number came out of him the way a man gives his age or his phone number, without inflection and without looking it up. It was September 2025, at Fortune's Brainstorm Tech conference in Park City, Utah, and Alyson Shontell, the magazine's editor-in-chief, had asked Eric Glyman about speed. Glyman was 35, the co-founder and CEO of a corporate card company called Ramp. Earlier that summer it had raised money at a $16 billion valuation. About eight weeks later it raised again, at $22.5 billion. It had just crossed $1 billion in annualized revenue. None of that was the number he chose to say.
"We're religious about it," Glyman said. "We count the days: We're 2,367 days old."
Shontell did a double take. "You know exactly how many days old Ramp is?"
"We do."
Asked why, he gave an answer that sounds, depending on your temperament, like either a monk's or a bookkeeper's: "Today is the only 2,367 we're going to have. We're going to make it count."
The arithmetic checks out. Eighteen months earlier, on the morning Ramp turned five, Glyman had posted that the company was 1,827 days old, a figure that includes one leap day. Count forward from that post and you arrive at 2,367 on almost exactly the week he walked onto the stage in Park City. Nobody in the audience was checking, and that is the point. The count is not kept for the audience. It is kept the way a casino keeps the drop, continuously and without sentiment, by people who understand that the only figure that matters is the one that accumulates.
Glyman grew up in Las Vegas, which may be why the comparison comes to mind. Las Vegas is a city organized around a single financial truth: the customer believes he is winning, and the house knows he isn't. Glyman built a company on the opposite premise, a credit card that makes money when its customers spend less. That is a strange thing to sell in an industry that has spent half a century teaching people that the purpose of a card is the points. What he sells is an alignment of incentives, and the sale has worked better than almost anyone expected. To see why, you have to go back to a rack of dress shirts in a Nevada mall, then to a bankruptcy advisory firm in Manhattan, then to an email sent on a Wednesday in August 2013, and finally to a stage in Utah where a man recites a number that only goes up.
By the Numbers
Ramp, Day by Day
2,367Days old when Glyman told Fortune's Brainstorm Tech audience, September 2025
$44BValuation after a $750 million round, June 2026
$1B+Annualized revenue, crossed in September 2025
70,000+Organizations using Ramp across North America and Europe
$12B+Customer savings Ramp reports, alongside 27.5 million hours
3%+Share of all U.S. corporate card transactions running on Ramp
45 / 60 / 70Days targeted for network approval, bank approval, and first funded transactions
The House and the Markdown
The Las Vegas Glyman grew up in did not hold still. In the eighteen years he lived there, as he has described it, the city went from a relatively quiet desert town to a metropolis of two million. Hotels went up, casinos expanded, and whole neighborhoods appeared in months. He was the middle of three brothers, a boy who loved math and watched a city scale before he had a word for scaling. The speed of it fascinated him. The economics of it, the casino as a machine for converting optimism into margin, he seems to have absorbed more slowly.
The education that mattered came at work. Glyman's first job was as a sales associate at Express, the clothing chain, in Las Vegas. That was 2008, the year he graduated as valedictorian of Green Valley High School. In the stockroom and on the floor he noticed something that bothered him. A dress shirt marked 20 percent off on a Tuesday could be 50 percent off the next day. In another telling of the same memory, he watched families buy clothes "on sale" and then saw the prices drop another 25 percent a few days later. Everyone at the register believed they had gotten a deal. The house, in a different costume, still won.
Most teenagers who notice this kind of thing conclude that retail is rigged and go back to folding sweaters. Glyman kept turning it over. The gap between what an item cost and what a customer paid, and the fact that the gap moved without explanation in the direction that favored the seller, became the question his career would keep returning to. It is a small question. He has built roughly $44 billion of enterprise value on it.
He went to Harvard and studied economics and East Asian studies, an unusual combination that sent him through the Harvard Beijing Academy, where he earned a citation in Mandarin in 2009. In 2011 he was a visiting scholar at Peking University, studying finance, accounting, health economics, and, oddly, anthropology. The anthropology fits better than it first appears. Much of his later work amounts to field observation of a particular tribe, the corporate finance department, and of its rituals: the expense report, the receipt chase, the month-end close.
He met Karim Atiyeh at Harvard. Atiyeh is the engineer of the pair, the one who would become Ramp's CTO and, much later, its co-CEO. In the early years he was also the one who, when the careful plans failed, tried something reckless and turned out to be right. Partnerships that last fifteen years usually depend on the two people wanting the same things and fearing different ones. The record suggests this one does.
After graduation in 2012, Glyman did not go to Goldman Sachs or McKinsey. He took a job as a financial analyst at Millstein & Co., where he stayed from July 2012 to September 2014. Ramp's own account describes it as a firm that helped struggling companies restructure their bankruptcies. It is an odd first job for a future founder and a fitting one. Restructuring is the part of finance that arrives after the optimism is gone, when the question is no longer how much a company can grow but how long it can live and who absorbs the losses. For two years the young analyst studied the end of the story before he had written any beginnings.
Ramp's in-house writers later described "extending the lifespan of businesses" as a recurring theme of his career. That is corporate-blog phrasing, but it holds up. The Las Vegas boy who noticed the markdown went on to spend two years among companies that had run out of runway.
A Refund for the Price of an Email
In 2013, Glyman and some friends booked a trip. The flights were locked in. The next day another friend decided to come and needed the same flight. When Glyman checked the fare, it had dropped substantially in the twelve hours since he paid. The difference was roughly $100. He thought it was a glitch and refreshed the page. It wasn't.
The Express stockroom had come back to him. The price had moved, and nobody was going to tell the customer.
He emailed Atiyeh that day, Wednesday, August 7, 2013, and the email survives, at least in the retelling Ramp produced for its fifth anniversary. The idea was a service that monitors price drops and secures refunds whenever prices fall. If the price of a flight drops, the app catches it and refunds you the difference. They met that night. They applied to Y Combinator. They didn't get in.
"Neither of us wanted to give up," Glyman says in that retelling. They decided to try again in six months. In the same clip he adds a line that sounds grandiose until you think about it: "We basically built the first AI agents." It is a 2024 vocabulary applied to a 2014 product, but it is not wrong. Paribus, the company they co-founded in January 2014, scanned a user's email for receipts, watched the prices of what they had bought, and when a retailer's price-adjustment policy entitled them to money, filed the claim on their behalf. The user did nothing. Software noticed the markdown and went to the counter for you.
The beta launched in September 2014. Glyman left Millstein the same month. Paribus debuted at TechCrunch Disrupt New York in May 2015, and that summer, on the second attempt, it went through Y Combinator.
Two episodes from that period still shape how Ramp operates, more than a decade later.
The first was a failure followed by a fluke. The team had set itself a goal of 10 percent growth and put in about 100 hours of work against it. They added 35 users in a week and missed the target. Then Atiyeh, essentially guessing, posted about Paribus on Reddit. Within twelve hours they had 400 new users. "It's hard to predict what exactly is gonna be the outcome," Glyman said years later, "and so just open yourself up to that asymmetry and try to find what really is more impactful — do more of that. Cut what isn't yielding those outcomes." The lesson he drew was not "post on Reddit." It was that effort and outcome are only loosely linked, and that the job is to keep measuring which inputs actually move things.
The second was a public dressing-down. YC held weekly office hours at which every startup reported three things: weekly growth rate, biggest problem, and what it was doing about the problem. Jessica Livingston, a Y Combinator cofounder, sat in on these sessions. That week Paribus had grown 20 percent. Its biggest problem was too many customer support tickets. Its proposed solution was to hire a customer support person, as the third employee of a two-person company.
Livingston took the proposal apart in front of the group. If the fix for support tickets was a hire, then next week, when they grew again, they would need another hire, and another after that. The real fix was to listen to customers and build a product good enough that they never needed to write in. She told them about Brian Chesky, already an icon at Airbnb, walking around with a Jawbone headset so he could take support calls himself. Glyman later called it "one of the best learning moments I've had as a founder," and wrote that she had schooled them "kindly and graciously," which is what people say after something humbling has turned out to be useful.
— Eric Glyman, on the lesson from Jessica Livingston at Y Combinator, summer 2015In a sense, every customer support ticket is a privilege – someone took time out of their day to write about how to improve your product. But it's also a failure – you could've saved them that time by making the product better and more intuitive in the first place. You can't out-hire a bad product, or compensate for poor taste with a big support team.
The coda is structural. "To this day, almost a decade later," Glyman wrote, "support reports into product at Ramp." The org chart still carries the mark of that meeting.
Paribus raised a seed round of about $2 million. It grew to around 700,000 users and worked with major retailers including Amazon, Best Buy, and Target. It also nearly died more than once. At one point, he told Harry Stebbings on 20VC, the company watched 80 percent of its revenue disappear overnight. He doesn't dramatize it. He says the team set aside its terror and worked through the crisis one step at a time. The fullest account of what that period felt like came later, in his description of the first-time founder's mind: "I was just always terrified and thinking, 'we're months away from running out of cash. We need to do everything by hand.'"
Then there was Amazon. Asked by Morning Brew to name the biggest mistake he would never repeat, Glyman went back to Paribus: "My first company, I didn't think about the long term. I treated Amazon like an enemy, not a partner. And I played a small part in getting Amazon to get rid of its price adjustment policy for every customer in America."
It is a striking confession, and it has a tragicomic shape. A boy who wanted to close the gap between what customers paid and what things cost built a company so good at exploiting a retailer's refund policy that the retailer eliminated the policy for everyone. He meant to help the shopper against the house, and the house responded by closing that game entirely. What he took from it was that adversarial cleverness has a ceiling, and that a business built on someone else's leniency survives only as long as the leniency does. He came away suspicious of any model in which his gain was someone else's loss, including the model he would soon see from the inside.
Capital One acquired Paribus in October 2016.
Inside the Machine That Sells Points
Glyman stayed at Capital One until February 2019, running Paribus as a senior director and leading what the bank called its savings automation team. The product became Capital One Shopping. By most measures this was a successful exit: a twenty-something founder selling to a top-ten American bank and keeping his product alive inside it. For the purposes of this story, its main value was the view it gave him of how the business worked.
From inside a major card issuer, Glyman saw something he found strange. "We saw what was a very profitable industry," he later said, "thinking a lot about how to get customers to spend more money to earn more points on the cards." The model, in his telling to David Senra, depended on persuading customers that points and rewards were valuable and then quietly devaluing them. It was the Express markdown again, at much larger scale. The customer believed he was winning, and the issuer's economics depended on him spending more so that he could believe it.
He has described the mechanics plainly, and the mechanics are why the insight was available to exploit. When a card is swiped, as he explained on one podcast, "the merchant gets the lion's share, processors take a little, Visa or MasterCard take a little, and the issuer keeps most of the interchange." Interchange, the slice taken from each transaction, is the issuer's economic engine. Every issuer therefore has a reason to want more swipes, and the rewards program is how that reason gets passed on to the customer. Spend more and earn more.
The customers, meanwhile, wanted something else. "As we talked to customers," Glyman said, "we saw and learned they were thinking not about how to earn more points or cash back, but were thinking a lot about how to get more in their bank account. And so we realized there was a large misalignment." Then he asked the question the rest of his career has been built on: "What if there was a card, and frankly software, that was designed not to get people to spend more, but ultimately to get those people to spend less?"
Phrased like that, it sounds like a slogan. In practice it is a bet that interchange from a customer who stays for a decade beats interchange from a customer who overspends for a year and then leaves resentful. Underneath that is a second bet: that a business whose customers spend less and grow faster will end up spending more in absolute terms, because it will be bigger.
The Paribus experience changed how he went about starting the next company. "Another big shift that happened as a second-time founder," he told Jack Altman and Erik Torenberg in 2023, "was doing a lot of work up front to validate the business model of Ramp, how it might work, figuring out the unique differentiation." Before launching a card, Glyman and Atiyeh talked with roughly 100 finance experts. What they heard was less about rewards than about drudgery: the waste of collecting receipts, coding transactions, and filing expense reports. The finance department was unhappy, and almost nobody in fintech was paying attention to it.
Tech:NYC later titled a profile of the company's early days with a phrase about going from "a sheet of paper" to a 13,000-customer business. The sheet of paper is a nice image, the business plan as a single page. Glyman's account of the early method is less romantic and more useful. "Many people think it's important just to start a company, go ship, and start selling it," he said in a Redpoint video. "But if you go and ask for a sale, you'll get feedback about why it is a fit, why it isn't a fit. If you ask for advice, say, 'I'd like to solve this problem for you and I'm building this product that I think will help you do that.' Simply ask them, 'Would this work for you? What are you looking for?'" Asking for a sale gets you a yes or a no. Asking for advice gets you the customer's actual problem.
In March 2019, Glyman, Atiyeh, and Gene Lee, a longtime friend, left to start Ramp. The incumbents they were taking on were old and large. American Express held about a third of the $2 trillion American corporate and small-business card market. Expense software belonged to Concur and Expensify. Ramp's founders were in their late twenties, carrying a product idea and a grievance about rewards. A South Park Commons promotion for one of Glyman's later talks put the generational contrast in a line: "Most of Ramp's competitors' founders wore top hats."
Forty-Five, Sixty, Seventy
Most founders hope to move fast. Ramp put the speed in writing.
"From early on we designed the company explicitly around velocity," Glyman has said. "Within 45 days we wanted network approval, within 60 days bank approval, within 70 days funding our first transactions." In fintech, the network is Visa or Mastercard and the bank is the sponsor that issues the card. Both are institutions whose internal calendars run in quarters. Committing to 45 days for one and 60 for the other was close to presumptuous. It was also a forcing function, because a deadline you announce to yourself rearranges every decision before it.
The deadline extended to the company's valuation. The goal, as Glyman has described it, was a $1 billion valuation within 18 months, a target they worked backward from milestone by milestone. "We wanted to go fast for sure," he said. "We wanted to go and make this company big and either make it huge quickly or fail really quickly." Note the symmetry he wanted. Failing quickly was an acceptable outcome. What he ruled out was the middle case, the startup that survives for years without amounting to much.
He was also, by then, the second-time founder with a different relationship to money. "I think the first-time founder mindset is very scarcity-focused," he told Altman. The second time, "once you've really found product-market fit, it's being able to lean into that growth and planning out phases ahead." At a pre-seed program for fintech founders called The Mint, he put it more bluntly: early on it is fine to watch every dollar, "as long as it doesn't build fear + prevent action," but don't be "penny-wise and pound-foolish." He also offered a line that sounds contrarian until you think about it: some of the most expensive hires become the cheapest over time.
Ramp launched formally in February 2020. Weeks later, COVID-19 shut down much of the country. For a company whose revenue depended on businesses swiping cards, launching into a lockdown looked fatal. Corporate travel collapsed, and with it a large share of corporate card volume. But the pitch Ramp was making, that you will spend less with us, suddenly matched the mood of every finance department in America. The timing that seemed catastrophic gave Ramp a market of CFOs who had just been told to cut costs.
The more consequential decision that winter was about software. Fortune's Leo Schwartz reconstructed it: "Competing with expense-report software like Concur and Expensify wasn't in Ramp's initial business plan, but the young team quickly realized that it was the natural next step. Rather than integrating their cards with another platform, why not build the software themselves?" The expense tool launched in February 2020 alongside the card. When an employee swiped, the expense was either processed automatically from the transaction data or the employee received a text asking for a photo of the receipt. "Goodbye, expense reports," Schwartz wrote. Then he gave the line that best describes the whole strategy: "If credit cards were the wedge for Ramp, expense reports were the mousetrap — the product that convinced customers to stick around."
The card got Ramp through the door. The software made leaving painful, because the finance team would have had to give up the hours the software saved them, hours they had stopped noticing they had. The mousetrap, notably, was another version of Paribus: software that handles the tedious part on the customer's behalf.
The numbers that followed became part of New York startup lore. Ramp reached a $1 billion valuation within two years of its 2019 founding, the fastest any New York startup had done it. At that point its revenue run rate was about $10 million. "Yes," Fortune's editors wrote, "that's quite a high multiple on revenue." In April 2021 it announced a $115 million Series B at $1.6 billion, after transaction volume rose about 400 percent in six months. By early 2022 it had passed $100 million in annualized revenue. Accounts of how long that took vary, from about seventeen months to eighteen to roughly three years, depending on where you start the clock. By any count it was among the fastest climbs in fintech.
The clock, at Ramp, starts in March 2019. The company keeps it running.
The Markdown, Again
In December 2021, at the peak of the zero-interest-rate era, Ramp was valued at $8.1 billion. The valuation was real in the narrow sense that investors had paid it. Ramp was barely three years old.
Then the Federal Reserve raised rates. The risk-free rate went from roughly zero to between 5 and 5.5 percent, and every discounted-cash-flow model in the venture world shrank, because a dollar of profit ten years out is worth much less when cash earns 5 percent in the meantime. Most companies from the 2021 vintage handled this by not looking. They raised no new money, or raised it with structure that preserved the headline number, and left employees holding options priced against a valuation that no longer existed.
Ramp did something else, and it is the decision that best reveals Glyman's character. The company went to the market, asked investors where they would price it, and accepted the answer. In August 2023 it raised $300 million at $5.8 billion, a reset of more than $2 billion below its peak. Glyman then stood in front of the company at an all-hands and walked every employee through the math: the discount rate, the DCF, why a company growing faster could be worth less on paper. Some employees were holding grants priced against $8.1 billion. He did not leave them to work it out themselves.
"In December 2021 we valued Ramp at 8.1 billion," he said later. "Today we actually think it's going to be a way bigger company and the value is 5.8 billion."
It is hard not to see the Express floor here. The shirt was 20 percent off on Tuesday and 50 percent off on Wednesday, and the shopper who paid Tuesday's price was never told. Glyman had spent his career objecting to undisclosed markdowns, to prices that moved against the customer in silence. When his own company's price moved against his own employees, he did the thing he had always wanted retailers to do. He announced the markdown at the register.
It was not a sacrifice for its own sake. An earlier piece of his advice, given on 20VC, reads differently in hindsight: never take the highest single round price. "Your goal is to maximize your ultimate enterprise value, not the value of a single round," Ramp's summary of the conversation reads. "Sometimes this means leaving capital on the table so you're able to pull together the right team and create more upside for your investors and employees." A valuation is a promise to the people who will hold the equity. Overstate it and you have sold them a shirt that will be cheaper on Wednesday.
The company grew 4x year over year in 2022. Then it kept going. In March 2025 Ramp hit $13 billion, after new and existing investors bought $150 million in secondary shares. In summer 2025 it hit $16 billion, then $22.5 billion eight weeks later, with $500 million raised. By November 2025 it was $32 billion. In June 2026, after a $750 million round, it was $44 billion. "We're back above all-time highs," he said in March 2025, back when $13 billion counted as a high.
He has since compressed his view of all this into four words. "Valuation is a derivative," he told the Brainstorm Tech audience. "Revenue comes from customers genuinely feeling that their trust was well-earned. When they signed up for a product, it delivered, and they told other businesses about it." A derivative in the financial sense: a price that depends on an underlying asset, here the customer's trust. He had seen what happens when the price comes loose from the asset, at Express, at Capital One, in the 2021 markup, and he does not seem to have much interest in watching it again.
Slope, Not Intercept
Asked by Morning Brew's Alex Lieberman for his single best hiring rule, Glyman answered in the language of a high school algebra class: "We hire for potential and growth trajectory and favor slope over intercept." The intercept is where a person starts, the résumé, the credentials, what they know today. The slope is how fast the line rises. "The goal is not an absence of problems, but 10x potential."
The phrase recurs throughout his public statements. In Park City he said it again: "Drive is about slope, not intercept. It's not just about what you know, but how quickly you can learn." By 2023 he was spending 35 to 40 percent of his own time on hiring and performance management, an unusual share for a CEO. He has said that a first-time founder treats this kind of attention as a luxury, and that a second-time founder learns it is the job. "Everyone is the hero of their own story," he told Altman. In his view, the reason people stay at a company is that they can see themselves growing there.
To find slope, Ramp looks for what one podcast summary called "asymmetric evidence of obsession": GitHub activity, leadership in niche communities, revenue streams people built for themselves. Ramp found engineers who had run profitable Minecraft servers at fifteen. That detail rhymes with Paribus, which also started as a side project that made money while its founders were supposed to be doing something else. Glyman prefers signals that are expensive to fake. He has argued that two days of working alongside someone yields more signal than fifteen hours of interviews, which is why referrals from people with firsthand knowledge carry weight.
His interviews for curiosity, the trait he says sits "about as close to the top of my Maslow's pyramid," follow a pattern any former analyst would recognize: "Tell me about a project that you feel was particularly successful. Why were you working on it? What happened? What were you specifically responsible for? For other people? Did it work? How'd you know it worked? ... Great. What would you have done differently, what worked beyond kind of your expectation and what went wrong?" The question that does the most work is the sixth one: How'd you know it worked? People who can't answer it have usually been measuring effort, and the Reddit episode taught Glyman not to confuse effort with results.
Keith Rabois, a Ramp board member, has a theory of early hiring he calls central casting: assembling the right chemistry, the way a producer casts a film. You hire people with sharp strengths, or "spikes," and then you hire the people they attract. Glyman uses the same word when he praises Fidji Simo. Simo is the French tech executive who took Instacart through its long-awaited 2023 IPO and then became the first CEO of Applications at OpenAI, reporting to Sam Altman. Glyman counts her among his closest advisors. "I just think that her ability to go and get the best out of people, to understand people's capabilities, to teach them and understand what are their spikes and help make their spikes sharper — I think that's extraordinary," he said. He called her "very underrated."
His other close advisor is Satya Nadella, whom he admires in terms that sound like a self-description. "Not only has he created something like a billion dollars in market cap for every day that he's been in the seat," Glyman said, "but I think what's more interesting about him is he is an avid listener. He's a reader. He's deeply curious. He wants to ask why." Notice that he measures Nadella's market-cap gains per day. Then the twist: Nadella saw AI coming first, Glyman argued, "because he was less busy talking and giving advice. He was more busy asking questions, being curious and reading." In 2023 both Simo and Nadella became individual investors and advisors to Ramp, along with Quora's Adam D'Angelo and Stanford AI Lab professor Chris Re.
Glyman doesn't use executive coaches. "I try to kind of go and call people up for an hour at a time," he said. He asks each person for advice on AI or marketing or sales, asks who they learned the most from, and moves on to that person: "just kind of jump from person to person." He takes his education in hour-long calls, one name leading to the next.
On firing, his rule is simple: "Ensure that if it's not working, it's never a surprise. Once you know, partner to make the transition smooth and gracious. How you treat those who put their trust in you matters." Again the rule is no hidden markdowns. On the fifth anniversary, writing about the 725 people then on staff, he added a line that sounds like a manifesto against the usual startup caste system: "Every role here is mission-critical; we have no second class citizens at Ramp." The same post mentions that building the company had at times meant "sleeping on the office floor or scrubbing toilets (done both)," and that the parenthetical is his.
Eleven Days Before Kickoff
When Ramp ran a Super Bowl ad in February 2025, the timing became part of the story. Most Super Bowl advertisers commit months ahead, pay the full rate, and hope. Ramp bought its slot eleven days before the game.
The late purchase was deliberate. By then the matchup was known, so Ramp could sign Saquon Barkley as a relevant, high-profile face for the spot instead of betting on an unknown. It also meant Ramp could pick up premium inventory abandoned by advertisers who had dropped out, including what Glyman describes as the final ad slot before kickoff, aired nationally. Some markets saw the ad twice. He declined to name the price but said it was lower than observers assume and the impact larger. February 2025 was Ramp's biggest February on record by a wide margin. Barkley's appearance on the Today show the following week, where the Ramp partnership came up, extended the campaign at no additional cost.
It was the Paribus instinct applied to advertising. Prices move. Someone who waits, watches, and is ready to act can buy the same thing for less than the people who committed early. He had once built software to catch markdowns on airfare. Now he caught one on the most expensive thirty seconds in American television.
A month later, announcing the $13 billion valuation, Glyman wrote the most revealing thing he has published, a LinkedIn post in which the self-deprecation and the ambition are hard to separate:
— Eric Glyman, announcing Ramp's $13 billion valuation, March 2025We're not Steve Jobs or Wilbur Wright. We won't invent the next iPhone or flying machine. Our job is more modest: save you time and money, so perhaps you can.
He then placed Ramp among the great efficiency stories of American business. "Sam Walton cut Walmart's supply chain down from 5 components to 3. Henry Ford cut the time to complete a car from 12 hours to 93 minutes. We cut the cost to buy anything from 30 minutes to 30 seconds." Comparing an expense-management product to the assembly line takes nerve, and a skeptic could fairly roll their eyes. But the claim underneath is reasonable and fairly old-fashioned. "Efficiency isn't flashy, but it builds enduring businesses," he wrote. "Growth gets the headlines, but efficiency leads to lasting progress." He quoted Charlie Munger: "Take a simple idea, and take it seriously."
The simple idea also turned out to be a good market position. Ramp estimated in March 2025 that it handled 1 to 2 percent of all American corporate and small-business card spend. Glyman's comment was "it's 98 to 99% to go," said with the cheerfulness of a man who has done that math more than once. His structural argument for why the pitch works in any economy is that cost-cutting math favors the company that cuts costs, whether the economy is booming or contracting. In a boom, savings are reinvested. In a recession, savings keep the company alive. Either way, a card that saves money has a customer.
Self-Driving Money
Roughly six months before ChatGPT launched, Ramp started using GPT-3. The use was unglamorous: classifying receipts and putting expenses in the right category. "Things that could be done deterministically with fuzzy logic," Glyman said, "but LLMs were uniquely suited." The large language model's first job at Ramp was bookkeeping.
When GPT-4 came out, Ramp's team built a Slack bot over a weekend and named it Toby. It ingested the company's sales calls. "No person has the time to listen to 50,000 sales calls," Glyman said, "but a large language model does." Anyone at Ramp, from an SDR to a product marketer, could ask Toby why customers chose Ramp over a given competitor and get an answer drawn from tens of thousands of conversations nobody could have reviewed. This was the Livingston lesson scaled up: listen to the customer, at a volume no person could manage.
What followed was a steady accumulation of measurable gains. Underwriting for a Ramp card used to take two days. After a two-week sprint in which the AI and underwriting teams sat side by side, it could take as little as four hours. With AI support, sales development reps booked up to four times as many meetings as the industry average. Glyman framed these gains carefully, as augmentation more than replacement: "Sometimes people tend to jump to full automation when what's really valuable in the short-run is augmenting what people are already doing." By late 2025, Ramp's policy agents were automatically approving 90 percent of expense transactions at 99 percent accuracy within the first month of use, and catching fifteen times more policy violations than manual review. Ramp had consumed more than a trillion tokens on OpenAI's models alone.
Glyman's name for where this is heading is "self-driving money." The phrase is evocative, but he means something specific. "Many tasks that need to be done manually right now will happen without you even asking," he told the anonymous writer behind CFO Secrets. "It will be the ultimate zero touch." He is skeptical of the rush to bolt chat interfaces onto everything. In his view, the most valuable AI is AI you never see, the kind that completes the expense report before you remember there was one. "Finance," he said, "is an exercise in systems design and systems thinking." He has also said, without much hedging, that AI could end the fifty-year American productivity slump, perhaps not in five years but possibly in ten.
That bet has changed how he describes his own competition. He no longer sees Ramp's real rivals as other fintechs. He sees them as AI labs, because what Ramp ultimately sells is automated knowledge work, not money movement. It has also changed what Ramp manages. The June 2026 announcement, Ramp at $44 billion, carried the title "The Third Pillar." The pitch fit in a sentence: "For 500 years, business spend was either on people or vendors. AI spending just created a third: intelligence." Payroll, vendors, and now tokens. Glyman noted to Senra that Anthropic and OpenAI were on track toward roughly $300 billion in combined annual revenue, around 1 percent of American GDP, and that most CFOs had no structured way to track it. "You could just hammer Salesforce all day," he observed, "you're not going to get a bigger bill from Salesforce." With models, every call costs something. Spend on intelligence behaves like spend on fuel, which makes it exactly the kind of variable cost a card built to reduce spending is suited to manage.
The company is no longer only American, and it is no longer run by one CEO. Ramp acquired Billhop, a Stockholm- and London-based payments platform, which gave it regulatory authorization in the UK and the EU. Its stated customer base now runs across North America and Europe and includes Visa, Uber, Shopify, and Anduril. A company that once relied on Visa for network approval now counts Visa as a customer. And Glyman gave Karim Atiyeh the same title he holds. "Today Karim Atiyeh and I are both taking new titles as Co-CEOs of Ramp," he wrote. "If you know us, this won't feel like a change." It reads as a formal acknowledgment of a partnership that began with an email in August 2013.
One more detail from this period helps explain everything else. Asked by Senra what he looks at when he thinks about design, Glyman recommended, of all things, a toaster. "Have you ever used a Breville toaster?" he asked. "If you care about design, you're meeting all these people, you should buy a Breville toaster. I find it very very interesting." What interested him was how it was made: "What they started doing is try to observe people toasting things."
That is the method in miniature. Watch the shopper at the register, the analyst at month-end, the employee holding a receipt. Watch long enough to see the friction the customer has stopped noticing, then remove it.
The Only 2,367
There is an obvious objection to counting the days, and Glyman has an answer to it.
The objection is that urgency becomes a cult. Companies that make a religion of speed tend to burn people out and mistake motion for progress. He has said nearly the opposite. "I would rather find people who just find extreme joy in their craft and set them up where they can be doing just that as much as possible all the time." In his telling, the day count is a measuring device, not a whip. "At the end of a seven day period, at the end of a month period, we look back and we ask, 'What did we get done over the past 30 days?'" he said in his Redpoint lesson. "Sometimes it's only a little bit." The purpose is to connect inputs to outcomes: "Where did I spend a lot of time that didn't drive outcomes?" Then cut those things and do more of what worked. It is the Reddit lesson, applied every month.
He has been open about where the urgency comes from. He credits Frank Slootman, the former CEO of Snowflake and ServiceNow, whose book Amp It Up argues that organizations slow down unless someone actively sets the tempo. Glyman's version is specific to the end of the work week. In a stagnating company, he said, it is easy on Friday afternoon to push a task to Monday. What a fast company needs is the urgency to finish it on Friday, which requires that "someone is driving and leaders are creating tempo."
He sets goals, but not the usual kind. "At Ramp, we avoid goals like 'grow revenue by X%,'" he wrote in 2024, "because it makes more sense to manage towards inputs, not outputs. Like if you want to win the Super Bowl, you don't practice winning the Super Bowl. You practice throwing, catching, running, tackling." The replies made clear the book behind the idea was Bill Walsh's. The outputs are not ignored. They are displayed. Ramp keeps a running scoreboard of the dollars and hours it has saved customers. It hangs on office walls, posts to the largest Slack channels, and can be queried through internal AI agents. "It's things that we talk about frankly every month," he said. He told Senra about Ken Griffin putting a handful of critical metrics on a permanent display at Citadel's headquarters where everyone could see them every day. The model is clear enough.
He applies the same discipline to himself. "Regular calendar audits," he answered when asked how he prioritizes. "They help confirm that how I actually use my time matches the company's top priorities. It's brutal but the reality check is critical." At Park City he described the larger version of the exercise, putting himself out of a job on a schedule. Compounding growth, he pointed out, works against you: what drove 100 percent growth last year might yield 50 percent next year and 25 percent after that. So every so often he asks: "what was I great at that I need to give up?"
On the morning of Day 1,827, Ramp's fifth birthday, Glyman wrote a post that reads, in retrospect, as the plainest statement of what he believes. "The secret is this: there is no secret to building a great company. It's a long, patient process with no shortcuts, no tricks, and often little glamor." He listed the things that had gotten them to day 1,000: hiring on potential, caring about customers, focusing on data, simply being kind. Those same things got them to day 1,827, he wrote, "and they'll be the same things that get us to day 10,000." The phrase day 10,000 is worth stopping on. It is about twenty-seven years. A man who counts in days had, without remarking on it, named a horizon close to a full career.
By the time of the third-pillar announcement, the count was somewhere past day 2,600. The scoreboard on the wall reported more than $12 billion and 27.5 million hours returned to customers. Unlike a price at Express, it is not marked down on Wednesday.
He signed off the anniversary post the way he signs off most things, with the next number.
"See you on Day 1828."
Part IIThe Playbook
What follows is an attempt to extract the operating logic from the story: the habits, structures, and decisions that recur across Paribus and Ramp often enough to be called a method. Some are familiar ideas carried further than most founders carry them. Others run against the standard startup advice of the last decade. All of them are grounded in what Glyman has said and done.
The principles
- 1.Find the incentive pointed the wrong way, and reverse it.
- 2.Count the days, not the quarters.
- 3.Write the deadline before you write the product.
- 4.Lead with a wedge, keep customers with a mousetrap.
- 5.Treat every support ticket as a product defect.
- 6.Mark yourself to market before the market does it for you.
- 7.Manage inputs, and put the outputs on the wall.
- 8.Hire for slope, and look for proof of obsession.
- 9.Make yourself redundant on a schedule.
- 10.Buy what others abandon, at the late price.
- 11.Build AI the customer never has to see.
- 12.Sell what lasts, not the current form of the product.
Principle 1
Find the incentive pointed the wrong way, and reverse it
Every incumbent industry has an incentive that quietly works against its customers. In corporate cards it was the rewards program, which taught businesses to spend more so the issuer could earn more interchange. Glyman found it by working inside the incumbent. His time at Capital One was less a reward for Paribus than an education in how the card business makes money. The question he came out with, "What if we wanted the same thing as our customers?", became the founding thesis. It also turned out to be a durable competitive advantage, because an incumbent cannot copy it without dismantling its own rewards economics.
The approach depends on noticing the misalignment and then believing it is worth building a company to fix. Glyman had been noticing it since the Express floor. The key move was converting a grievance into a business model: a customer who saves money stays, refers others, and grows, and the card earns its modest interchange on a larger, longer-lived pool of spending. By 2026 the typical Ramp customer, he says, cuts expenses by more than 5 percent a year while growing revenue at a median of 16 percent, against a national average of 3 to 4 percent.
There is a defensive lesson too, and it comes from Paribus's mistake. Exploiting the incumbent's leniency, as Paribus did with Amazon's price-adjustment policy, works until the incumbent closes the loophole. Aligning with the customer is something the incumbent can't take away.
Tactic: List the three ways your industry's dominant players profit when their customers do worse, and build your pricing so that your revenue grows only when those customers do better.
Principle 2
Count the days, not the quarters
The day count is easy to dismiss as a gimmick, and it would be one if it were only a slogan. At Ramp it functions as a unit of measurement. Days are small enough that wasting one feels real. Quarters are large enough that drift goes unnoticed. "Today is the only 2,367 we're going to have" is a sentence about scarcity, but not the fearful scarcity Glyman disavows from his first-time-founder years. It is a reminder that time, not cash, is the resource that can't be replaced.
The count is paired with a review cycle. Every seven and thirty days the question is the same: what got done, which inputs drove it, where did effort go without result. The Slootman influence is the Friday rule, where work that could slip to Monday gets finished Friday because a leader is setting the pace. The day count gives that pace a visible measure.
Glyman's explicit claim is that this is not urgency for its own sake. The short window exists to make tradeoffs visible, so that work which "seemed useful" but didn't move anything can be cut, and the work that did can be expanded.
Tactic: Put your company's age in days at the top of your weekly review, and require every team to name one piece of work from the last 30 days that it will stop doing.
Principle 3
Write the deadline before you write the product
Most founders set goals they hope to hit. Ramp set deadlines it designed around: network approval in 45 days, bank approval in 60, first funded transactions in 70, and a billion-dollar valuation in 18 months. These numbers weren't forecasts. They were constraints, meant to remove options and force decisions that would otherwise drag on for months. "Either make it huge quickly or fail really quickly" is a willingness to accept a quick failure in exchange for avoiding a slow, ambiguous middle.
The same logic governs financing. Glyman's advice from 20VC is to treat every round as a science experiment: what must this capital prove to unlock the next stage? That framing turns money from a cushion into a hypothesis with a deadline. It also explains why he warns against taking the highest price, since an overpriced round raises the bar for the next experiment.
He also distinguishes between decisions that deserve speed and decisions that deserve care. "When it comes to core foundational building blocks, take the time to get them right. For all other issues, what matters most is iterating fast." The 100 interviews with finance experts before launch were the slow part. Everything after them was fast.
Tactic: Before your next initiative, write down the date by which it must show evidence or be killed, and share that date with someone who will hold you to it.
Principle 4
Lead with a wedge, keep customers with a mousetrap
The card was never the whole business. Leo Schwartz's line, that cards were the wedge and expense reports the mousetrap, describes a two-product strategy that Ramp arrived at before most startups have finished a first product. The wedge is easy to try and easy to explain: a corporate card. The mousetrap is the software that, once in place, makes leaving costly because it removes work the customer no longer remembers doing.
The decision to build rather than integrate is the important part. Concur and Expensify already existed, and the default would have been to plug into them. Building the expense layer in-house gave Ramp the transaction data, the user relationship, and the ability to keep extending: bill payments, procurement, travel, treasury, accounting automation, and eventually AI spend management.
R
Wedge, Mousetrap, Platform
How each Ramp product layer extends what the previous one did.
| Layer | What it does for the customer | Role |
|---|---|---|
| Corporate card | Spend that is designed to fall, not rise | Wedge |
| Expense automation | Receipts matched by text, reports filed automatically | Mousetrap |
| Bill pay, procurement, accounting | One place for spend, approval, and the books | Platform |
| AI spend management | Tracking the "third pillar" of intelligence spend | Platform |
Tactic: Identify the most tedious task your customer performs immediately after using your core product, and build that next, instead of integrating with whoever currently owns it.
Principle 5
Treat every support ticket as a product defect
Jessica Livingston's critique of Paribus's plan to hire a support person is now built into Ramp's structure. Support reports into product. That reporting line expresses a belief: support volume is a symptom, and a support team, however good, can't fix the cause. "You can't out-hire a bad product."
Glyman's reframing is that each ticket is both a gift and an indictment. Someone spent time telling you how to improve, and that time should never have been necessary. The metric he points to is the gap between active-user growth and support-ticket growth. Users should grow much faster than tickets. If tickets grow in step with users, the product is creating its own workload.
Toby, the Slack bot that listened to 50,000 sales calls, is the same principle extended to sales. The Chesky headset became a language model. The goal is still to hear the customer directly, without layers of summary in between.
Tactic: Move your support function's reporting line to product, and track the ratio of support tickets to active users as a core product metric.
Principle 6
Mark yourself to market before the market does it for you
The $8.1 billion to $5.8 billion reset is the most unusual decision in Ramp's history and probably the most useful one to study. When rates went from near zero to 5 percent, Glyman did not protect the headline number. He went to the market, accepted the lower price, and explained the discount-rate math to every employee at an all-hands.
Two things made this work. First, it was transparent: employees learned the real value of their equity from the CEO rather than discovering it later. Second, it removed a stale number that would otherwise have hung over every subsequent negotiation. A company valued at a fictional price must either grow into it or eventually reset under worse conditions. Ramp reset early and then outgrew it, reaching $13 billion by March 2025.
The habit underneath is the same one Paribus was built on, intolerance for a price that has silently drifted from reality.
$
The Valuation Path
Raising, resetting, and outgrowing.
2019
Ramp founded in March; the day count begins.
2021
$115M Series B at $1.6B in April; peak of $8.1B in December.
2023
Self-imposed reset: $300M raised at $5.8B, explained to every employee.
2025
$13B in March; $16B and then $22.5B over the summer; $1B annualized revenue; $32B in November.
2026
$750M round at $44B; Billhop acquisition opens Europe; Glyman and Atiyeh become co-CEOs.
Tactic: If the last price on your equity no longer reflects current conditions, reprice it yourself and explain the math to your team before an investor or the market does it for you.
Principle 7
Manage inputs, and put the outputs on the wall
Glyman's Super Bowl analogy, that you practice throwing and tackling rather than "winning," is a familiar management idea. Ramp's version has two parts. Goals are set on inputs: listening to customers, sales craft, and shipping throughput. The outputs that justify those inputs, dollars and hours saved for customers, are displayed everywhere, on office walls, in the largest Slack channels, and through queryable agents.
This design answers the most common objection to input management, which a commenter raised under his post: input goals can shield a team from accountability for results. The scoreboard prevents that. Everyone manages inputs, and everyone can see whether the inputs are working. The Ken Griffin story Glyman tells, a few critical metrics on a permanent display, describes the same idea: a small number of visible measures, not a dashboard of hundreds.
The choice of output matters as much as its visibility. Ramp's scoreboard measures customer outcomes, not Ramp's revenue. "Valuation is a derivative" applies here too. Revenue follows from customer savings, so customer savings is what goes on the wall.
Tactic: Choose one customer-outcome metric, display it where every employee sees it daily, and set team goals only on the inputs you believe move it.
Principle 8
Hire for slope, and look for proof of obsession
"Slope over intercept" is Glyman's most repeated idea about people, and he backs it with specific methods. Ramp looks for evidence that is hard to fake: GitHub activity, leadership in niche communities, self-built revenue streams, or teenage Minecraft servers that made money. It prefers two days of working together to fifteen hours of interviews. It runs a structured conversation about a past success that ends with how did you know it worked? and what went wrong?
The underlying belief is that capability compounds when a curious person sits in a fast-moving environment, and that the most expensive early hires can be the cheapest in the long run. Glyman's 35 to 40 percent time allocation to hiring and performance management backs up the rhetoric, and so does his practice of monthly development conversations, distinct from weekly one-on-ones, about how a person is growing.
The firing rule completes it: never a surprise, always gracious. A system that bets on potential will sometimes be wrong, and how those exits are handled determines whether people keep trusting the system.
Tactic: In your next interview loop, replace one credential-based screen with a paid working session, and ask every finalist how they knew their best project actually worked.
Principle 9
Make yourself redundant on a schedule
The arithmetic of compounding is Glyman's argument for continual self-replacement: what produced 100 percent growth last year produces 50 percent next year and 25 percent the year after. A CEO who keeps doing what made them good becomes the constraint. So he runs regular calendar audits ("brutal but the reality check is critical") and periodically asks himself a hard question: "what was I great at that I need to give up?"
The other half is how he learns. He doesn't hire a coach. He calls practitioners, Fidji Simo, Satya Nadella, and others, for an hour at a time, asks each one who taught them, and follows that lead. It is a fast, inexpensive way to learn from people who have already solved the problem.
The co-CEO announcement can be read the same way. Glyman shared the title formally instead of guarding it. "If you know us, this won't feel like a change" suggests that the work had already been split, and the titles caught up.
Tactic: Once a quarter, audit your calendar against your company's top three priorities and hand off one responsibility you are still good at but no longer need to own.
Principle 10
Buy what others abandon, at the late price
The Super Bowl purchase, eleven days out, a matchup-relevant star, the final slot before kickoff, shows a pattern older than Ramp. Prices move, and the patient buyer who stays ready can pick up premium assets that early buyers paid full price for or gave up. Paribus was software built to catch that kind of markdown. The Super Bowl buy applied the same instinct to media.
The approach requires two things most companies lack. One is the operational speed to act on short notice: creative, talent, and approval within days. The other is the confidence to wait while competitors lock in. Ramp's velocity culture provided the first. The second came from recognizing that in markets with fixed supply and dropouts, the late price is often the better one.
The principle applies beyond advertising: talent freed up by a competitor's layoffs, office space in a weak market, equipment from a failed company. Being prepared to move fast lets you be the buyer when sellers need to sell.
Tactic: Keep a short list of high-cost assets your company wants, and maintain the approvals and readiness to buy any of them within ten days when the price drops.
Principle 11
Build AI the customer never has to see
While competitors added chat windows to their products after ChatGPT, Glyman argued for "zero-touch" automation: AI that works in the background and finishes the task before anyone asks. The expense report that completes itself, the policy agent that approves 90 percent of transactions at 99 percent accuracy, the receipt classifier that ran on GPT-3 before most people had heard of language models. "Self-driving money" means software that does the work instead of a chatbot that talks about it.
Two corollaries come with it. First, lead with the benefit, not the technology. Ramp sells time and money saved, not "AI-powered finance." Second, augmentation often beats full automation in the short run. The underwriting sprint that cut approval time from two days to four hours worked because the AI team sat beside the underwriters. It didn't replace them.
There is a cost side as well. Glyman benchmarks models and routes most tasks to cheaper ones and only a small share to expensive frontier models. Treating intelligence as a variable cost to optimize is consistent with a company built to reduce its customers' spending.
Tactic: For each AI feature on your roadmap, ask whether the user would prefer never to see it, and if so, build the version that runs without being asked.
Principle 12
Sell what lasts, not the current form of the product
In his April 2026 conversation with Patrick O'Shaughnessy and David Senra, Glyman used American Express as his example. In his telling, it began by moving money safely across a country whose banks couldn't be trusted, and it now sells corporate cards and restaurant reservations. What it sold all along was trust. "Great companies often understand they are not selling a product," he said. "They are selling something enduring." His test question for founders: "Even if the form factor of the product changes, will the service you're selling endure?"
For Ramp, the enduring service is save time and money. The card was its first form, expense automation its second, and AI spend management its latest. This is why Glyman now names AI labs, not fintechs, as Ramp's true competitive benchmark. If what lasts is automated knowledge work, the competitors are whoever automates knowledge work best.
The principle also guards against a common failure: defending a product after its moment has passed. A company that identifies with a card will fight to protect the card. A company that identifies with saving time and money will drop the card the day something better arrives.
E
Form Factor vs. What Lasts
Where the product has changed and where it hasn't.
| Era | Form factor | What it sold |
|---|---|---|
| Paribus, 2014–2016 | Email-scanning refund agent | Money back on purchases already made |
| Ramp, 2019–2020 | Corporate card plus expense software | Less spend, fewer hours |
| Ramp, 2026 | Financial operations platform plus AI agents | Less spend, fewer hours, including spend on intelligence |
Tactic: Write one sentence describing what your company sells that would still be true if your current product disappeared, and test every roadmap item against it.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Eric Glyman, Fortune Brainstorm Tech, September 2025We're religious about it. We count the days: We're 2,367 days old.
— Eric Glyman, LinkedIn, on Ramp's fifth anniversaryThe secret is this: there is no secret to building a great company. It's a long, patient process with no shortcuts, no tricks, and often little glamor.
— Eric Glyman, Fortune Brainstorm Tech, September 2025Valuation is a derivative. Revenue comes from customers genuinely feeling that their trust was well-earned. When they signed up for a product, it delivered, and they told other businesses about it.
— Eric Glyman, on his biggest mistake, to Morning BrewMy first company, I didn't think about the long term. I treated Amazon like an enemy, not a partner.
— Eric Glyman, with David Senra and Patrick O'Shaughnessy, April 2026Great companies often understand they are not selling a product. They are selling something enduring.
Maxims
- Find the incentive pointed the wrong way. The easiest market to take is one where the incumbent profits when its customers do worse.
- Count in days. A quarter hides drift that a day makes obvious.
- Write the deadline first. Forty-five days to network approval rules out everything that would have taken ninety.
- A support ticket is a product defect. You can't hire your way out of a product that confuses people.
- Announce your own markdown. Reprice honestly and early, before someone else reprices you and your people find out last.
- Measure customer outcomes, not your own revenue. Put the dollars and hours you save customers on the wall, and the revenue tends to follow.
- Slope beats intercept. Hire for the rate of learning, and look for proof that is hard to fake.
- Make yourself redundant on purpose. Regularly ask what you were great at that you now need to give up.
- Buy at the late price. Speed lets you pick up the premium assets others abandoned.
- The best automation is invisible. Build software that finishes the work before anyone has to ask for it.
