The expense report is a peculiar relic. A company lets money leave the building, then asks somebody to explain it days or weeks later. The employee searches for a receipt. A manager remembers the project. Finance determines the account. Software moves the same facts from one box to another, and everyone congratulates the software for making the chase slightly faster. Oded Zehavi has spent most of his career close enough to payment rails to see the absurdity. At Mesh, the company he co-founded in 2018, he is working toward a different order of events: permission, policy and purpose travel with the payment from the beginning.
Zehavi's favored description of himself is less grand than his résumé. He has called himself a “payment plumber.” It fits. Before Mesh, he helped PayPal launch services across the Middle East and Africa, ran revenue at Payoneer, and served as chief operating officer at mobile-security company Kaymera Technologies. His undergraduate degree, earned at the Hebrew University of Jerusalem from 1986 to 1989, is in computer science. Across the years, the job titles changed from technical to commercial to operational. The subject stayed stubbornly consistent: how money moves, and what businesses must know while it does.
“I, in some ways, position myself as a payment plumber.”Oded Zehavi, describing nearly 30 years in payments
A market disappears
Mesh began a year before COVID-19 with a different idea. Zehavi has said the company was building a product for a world that the pandemic promptly switched off. Customers disappeared; the product no longer fit the new conditions. This is where founder biographies often acquire a suspiciously tidy hinge: the obstacle arrives, insight flashes, success follows. Zehavi's own account is more useful because it keeps the unpleasant bit. The company found itself without customers. Its team and technology remained, but their intended destination did not.
The team looked at what the pandemic had made newly urgent. Work had scattered into homes and across borders. A finance department could no longer lean over a desk to ask about a charge. Company purchasing sprawled across software subscriptions, home-office orders and travel that would eventually return in altered form. Mesh turned toward spend management for distributed companies. The payment remained important, but it became one piece in a larger machine of approvals, controls, receipts and accounting data.
By the time Mesh announced a $60 million Series C in September 2022, it reported more than 1,000 customers, close to $1 billion in annualized payment volume, and roughly 150 employees in Israel and the United States. Total funding reached $123 million. Zehavi stressed discipline rather than theatrical valuation. Mesh said its revenue run rate had tripled during the first half of that year; the new money would support research, development and go-to-market work. The announcement landed during an economic downturn, when “growth at any price” was losing its charm and every expense had begun to deserve cross-examination.
The transaction is only half the story
A card network is very good at recording that a purchase happened. It is less gifted at explaining why. The merchant and amount appear immediately; the trip, client, budget, department, policy exception and general-ledger code live elsewhere. Traditional expense management begins after this context has scattered. Employees type it back into existence. Finance checks the reconstruction. Managers approve a purchase that has already occurred.
Mesh's product argument is that the sequence should be reversed. A virtual or physical card can carry rules before the swipe. A receipt can be collected and matched automatically. The accounting system can receive coded information. Managers can see spending as it happens, not as an archaeological report at month's end. The ambition is not simply a better corporate card. It is a transaction with a memory.
This helps explain why Zehavi's language has migrated from payments to automation and, lately, to “invisible T&E.” Travel and expense work is full of visible chores because systems fail to share context. A booking sits in one platform, a calendar invite in another, a card charge on a network, an emailed invoice in an inbox, and policy in company software. The employee becomes the integration layer, which is a rather expensive use of a human being.
The purchase that explains itself
trip or project
before spend
with controls
matched live
ready to close
Orchestration, not conquest
In May 2026, Mesh introduced what it calls an Orchestration Layer. It is designed to bring together data from booking tools, payment networks and other travel activity. The company said it connected with established booking platforms and transactions across Visa, Mastercard and American Express, as well as Mesh's own payment product. Calendar invitations, emailed itineraries and invoices can contribute the facts that conventional systems leave stranded.
The strategy matters because global companies rarely run on one immaculate stack. A useful travel partner in Canada may not be the right one in India. A local card program may solve a regulatory or currency need that a global mandate cannot. Zehavi has summarized what enterprises tell Mesh: they already have partners they trust and do not want software to force a replacement. Mesh's answer is to sit beneath the assortment, gathering the story without demanding that every country work identically.
This is an operator's response to enterprise complexity. The easy diagram has one vendor. The real company has acquisitions, regional habits, old contracts, favorite agencies and employees who book flights wherever they can find them. Orchestration concedes the mess and tries to make it legible. It is less conquest than diplomacy, conducted by APIs.
“A great leader is a person that people will follow over and over again.”Oded Zehavi on leadership
The repeat-follower test
Zehavi's test for leadership is equally practical. A leader, he has said, is someone people will follow repeatedly; one performance can fool an audience, but repetition reveals the real thing. A former colleague's public recommendation describes him as a manager able to see the big picture while discussing the finest details of the underlying technology. The two observations belong together. Payment businesses punish leaders who float too far above the pipes. They also punish those who stare at a valve and forget the building.
His route to founding supports the point. At PayPal he worked on a regional expansion playbook. At Payoneer he oversaw sales, marketing and product. At Kaymera he ran operations. He has also served on boards or advisory boards connected to AU10TIX, Rewire, Fiverr and CreditStacks. Each role offered a different view of how a financial-technology company crosses borders, earns trust and survives the distance between a demo and daily use.
There is personality in the way he discusses the work. He is fond of plumbing, playbooks and systems rather than founder mythology. He talks about customers and colleagues by name. He is a football fan. When Mesh's funding news appeared on the Nasdaq tower in 2022, his public reflection focused on the team and the strange distance traveled since launch. The pleasure was visible; so was his awareness that a favorable round in a difficult market was luck as well as labor.
When the interface recedes
The latest version of the thesis reaches beyond automated forms. In 2026, Zehavi began speaking publicly about agentic commerce and AI assistants connected to live spend data. Mesh announced an MCP connection intended to let a user ask a plain-language question of company spending from tools such as ChatGPT, Claude or Gemini. The practical promise is modest in wording and substantial in implication: fewer dashboards to patrol, fewer reports to assemble, and answers drawn from data already in motion.
There are serious questions beneath the convenience. When software agents can read spending, apply policy or initiate purchasing decisions, accountability and security must be designed into the pipes. Zehavi joined a New York Fintech Week panel on open banking security and accountability in 2026, and Mesh's invitation to appear as an agentic-commerce partner at Visa's Payments Forum placed the company close to the payment networks that will help decide those rules.
Still, the durable idea is not AI. It is context. A system becomes useful when it understands enough about a purchase to spare a person from explaining the obvious. The ideal expense product may therefore look like no product at all: a policy applied quietly, a receipt found automatically, a ledger updated correctly, a finance team free to notice what the numbers mean.
Zehavi's career has followed payments from geographic expansion to platform scale to corporate automation. The destination he now describes is an absence - fewer forms, fewer chases, fewer tiny acts of clerical penance. It is a funny ambition for a software founder. Most products want attention. The best plumbing is noticed only when it fails. If Mesh succeeds on Zehavi's terms, the compliment will be silence.