Erez Agmon Wants AI Agents to Run Revenue, and He’s Building the Infrastructure to Let Them

The Fintech Founder Building the Revenue Infrastructure and intelligent layer for the AI Era
Erez Agmon has a phrase for what he found inside most software companies’ finance departments: “the world’s busiest spreadsheet.”

The line belongs to one of his customers, Narmi’s VP of finance, but Agmon repeats it because it describes the problem his company was built to solve. Pricing in software has changed faster than the systems built to bill for it. It’s no longer just usage. Companies are now selling the work their AI agents do: resolving tickets, reviewing documents, closing out tasks that used to be billed as service hours. That shift brings services-style contract complexity into software billing: milestones, delivery phases, prepaid credits and top-ups, step-up commitments and success-based fees, often all in the same agreement. Every new plan, usage tier or contract exception meant an engineering ticket. Every month-end meant a finance team reconciling usage data, contracts and invoices by hand, trying to work out what had been delivered, what could be billed and what could be recognized.

Then AI changed the economics underneath it all.

“AI broke the old math of software,” Agmon says. “A SaaS seat used to cost almost nothing to serve. Today a single heavy customer can wipe out the margin on an entire contract, and most finance teams find out a quarter later.”

Vayu, the company he co-founded in 2024, is his answer: an AI revenue management platform built in three layers. At the base is billing infrastructure that captures and prices every billable event and contract term. On top of it, AI agents run the revenue process end to end, from billing through revenue recognition. Above both, an intelligence layer tells a CFO where margin is being made and where it’s leaking. Finance teams at Vi, Groundcover, Narmi, Simetrik, Dataplor, Aquant, Zesty and Solidus Labs run it. Vayu has raised more than $10M from Flint Capital and The Garage, with participation from the founders of Melio and former managing partners at SoftBank. The company is headquartered in New York with R&D in Tel Aviv.

Above it all: intelligence layer that protects margin

The newest layer is the Revenue Intelligence Hub. There, finance leaders can ask Vayu’s Insights Agent plain-language questions about usage, contracts, pricing, revenue and margins, and get answers grounded in live data. It also surfaces what finance should act on before month-end: customers whose consumption is outpacing their plan, prepaid credits running low ahead of a top-up, pricing tiers where cost-to-serve is eroding margin, revenue leaking between contract terms and invoices, and renewals at risk.

This is where Agmon’s argument about AI economics comes together. “The next phase of enterprise AI will be measured by the cost of completed work, not the cost of a token,” he has said. For a company selling agent-delivered work, that means knowing what every customer pays against what it costs to serve them. For finance, completed work means an invoice that went out correctly, revenue recognized properly and every payment reconciled, not a summary of what someone should do next.

From investor to operator

Agmon spent years on the investing side, at Fresh.Fund and in early-stage deals, before joining the founding team of the payments startup PayEm. He holds finance and law degrees from Reichman University, sits on the Forbes Finance Council and is a member of the Operators Guild.

The investing years taught him to see patterns across companies rather than problems inside one. The pattern he kept seeing was pricing breaking loose from billing. software moved from flat subscriptions to usage, credits and hybrid models. Then AI agents turned software into services, and billing inherited all the complexity of services contracts. Every contract term changes what a company can bill and when it can recognize the revenue. And every AI interaction behind it carries a real cost in tokens, compute and third-party models.

“Companies were launching pricing experiments faster than their billing systems could handle them,” Agmon says. “Billing was built for a subscription that renews on the first of the month. That world is gone. The inputs now move every day, and the infrastructure has to move with them.”

The foundation: billing infrastructure built for modern pricing

Vayu started by rebuilding the layer everything else depends on. Its billing infrastructure gives engineering teams a single integration point for usage data. It ingests and validates up to one million usage events a day and turns them into billable, priced records. It unifies that usage with contracts, CRM and financial data from Snowflake, AWS, Azure, Salesforce, Stripe, NetSuite and others. It supports usage, credit, hybrid and outcome-based pricing without custom code, along with the contract structures that come with agent-delivered work: milestones, phases, prepaid credits and top-ups, step-ups and ramps.

The design goal is that developers connect once and are then out of the loop. New plans and pricing changes stop being engineering tickets and become something finance launches itself. Customers are typically live in 14 days or less.

On top: agents that run the revenue process

With clean, priced data underneath, Vayu’s AI agents handle the revenue process end to end. Contract extraction turns signed agreements, however complex, into billing and pricing logic. Billing agents generate and send invoices, whether they are triggered by usage, a completed milestone or a credit top-up. Revenue recognition agents apply ASC 606 rules to every contract, including the timing of milestones, delivery phases, prepaid credits and step-up commitments, and produce audit-ready reporting. Collections and reconciliation agents match every payment back to the invoice and the ledger.

“The agents handle what is routine,” Agmon says. “The people handle what is not. That is the whole design. Nobody wants an AI approving an unusual invoice on its own, and nobody wants a person re-keying a thousand routine ones.”

The company reports 75 percent faster billing, 90 percent faster reconciliation and zero spreadsheet errors across its base, and one customer cut its monthly close from 15 days to three or four.

The report that named the problem

In 2026 Vayu published the CFO Signal Report with PwC and The SaaS CFO, surveying finance leaders at B2B software companies on pricing, revenue operations and AI. Its headline: the revenue engine is breaking under modern pricing. [Insert two or three headline statistics, ideally one on margin or cost visibility.]

Agmon has taken the findings into rooms he organizes himself: invitation-only dinners for finance, revenue and operations leaders in New York, with Los Angeles, San Francisco, Denver and Miami next. There are no panels and no pitches. “The conversation always comes back to the same question,” he wrote after one of them. “How can finance teams move faster as pricing, usage and revenue models become more dynamic, without losing trust, context or control?”

Lessons for founders

Agmon has three pieces of advice for anyone building in finance software.

Build the foundation before the features. Agents are only as good as the data under them. Vayu invested first in billing infrastructure and a unified revenue data layer, because an agent reading from a broken spreadsheet just produces errors faster.

Start where the complexity is. Every customer’s pricing is different, and as agents take on services work, the contracts keep getting stranger. That difficulty is exactly why billing was left behind, and exactly why it’s a moat for whoever solves it.

Sell to the person who owns the outcome. Vayu sells to CFOs, not engineering, because the finance leader is the one who has to explain the number, and the margin, to the board.

What comes next

Agmon wants revenue to be explainable before month-end rather than after. He wants a CFO to see at any moment what the company has billed, what it has earned, what it cost to deliver and why, and to act on it by launching a pricing change, catching a leak or repricing an unprofitable customer rather than reporting on it weeks later.

“Every company is going to run some version of usage or outcome pricing,” he says. “The ones that win are the ones whose revenue systems can keep up, and who know which revenue is actually profitable. We want to be the reason they can.”

It is an ambitious goal described modestly: turning billing from the part of finance everyone works around into the infrastructure everything else runs on.