Ramp CEO and cofounder Eric Glyman sat down at Fortune’s Brainstorm Tech conference to break down the company’s rapid ascent and operating philosophy. Launched in 2019, Ramp became the fastest New York startup to reach unicorn status, a trajectory Glyman credits to an obsessive focus on speed and alignment with customer outcomes. He says Ramp doubled revenue over the last year while generating more cash flow than the prior year, and the team “counts the days,” 2,367 at the time of the interview, to instill urgency and prioritize the work that truly moves the business forward. Rather than pushing points and higher spend, Ramp’s mission flips traditional corporate card incentives by helping customers spend less and save time, and the company ships features at a pace that exceeds the number of business days.
Glyman argues that first-mover advantage was less important than building a product tightly aligned with customer value and iterating at Valley speed in a financial services market long starved of innovation. He frames the company’s growth as the product of this bias for action, constant measurement, and a culture that forces explicit trade-offs. Ramp supports more than 45,000 companies, from family farms to the Fortune 500, and invests heavily in product, with more than half of payroll devoted to R&D, engineering, data science, and design.
Artificial Intelligence underpins many of Ramp’s workflows. Tap a card, snap a receipt from a text prompt, and Ramp automatically matches the image to the transaction and completes accounting categories, collapsing expense entry to roughly 10 seconds. The company also uses Artificial Intelligence to fight Artificial Intelligence. As AI-generated receipts proliferate, Ramp partnered with OpenAI, Anthropic, and others to build detection systems, drawing on a repository of more than 100 million receipts and cross checking multiple sources of truth such as card data, merchant data, images, and accounting records. Glyman says this multi signal approach is more robust than systems that rely on images alone.
Ramp’s policy agents, powered by large language models, read a customer’s expense policies in depth and auto approve about 90 percent of transactions, flagging the remaining 5 to 10 percent with explanations. Glyman says the system is roughly 99 percent accurate, which he describes as around ten times more accurate than the average employee, saving managers significant review time while catching issues that would otherwise slip through. Since inception, the company has automated 27.5 million hours of work and helps the average customer cut annual expenses by more than 5 percent. He adds that Ramp measures and reports ROI rather than selling technology for its own sake, pointing to a net promoter score in the sixties, comparable to Apple.
On leadership, Glyman says he aims to “put himself out of a job” by delegating to specialists and focusing on the few areas where his time has the highest return. He relies on mentors including Fidji Simo, now at OpenAI, and Microsoft’s Satya Nadella, and believes most finance work will shift from backward looking reconciliation to forward looking capital allocation as automation progresses. Artificial Intelligence is not replacing CFOs, he says, but it is eliminating low level tasks like expense reports and categorization so people can spend more time on the work that matters.
