Ofer Feldman

One Invoice at a Time

Somewhere in Mountain View, a finance team approves an invoice in seconds that used to take days. They probably don't think much about it. Ofer Feldman does. As Co-Founder and CTO of Stampli, he has spent a decade making that mundane moment faster, smarter, and more reliable - and the compound result is a business processing over $100 billion in invoices annually for more than 1,600 companies worldwide.

Stampli is not a household name. It operates in a part of the enterprise stack that rarely makes headlines: accounts payable. But if you've worked in finance at a mid-to-large company, you've felt the problem Ofer is solving. Invoices pile up. Approvals stall. ERPs don't talk to each other. Payment runs get delayed. The backroom of B2B commerce is chaotic in ways that spreadsheets and email have never adequately fixed.

Ofer's answer was not a workflow tool. It was an AI - built carefully, trained on real data, deployed into real accounting systems. Billy the Bot, Stampli's AI assistant, launched in 2015 and has been learning from every invoice it has touched since. That is nearly ten years of continuous training data that no competitor building "AI-first" features in 2023 can replicate overnight.

That is a competitive moat disguised as a patience argument. While others demo AI, Stampli measures it. Ofer's philosophy is strikingly pragmatic: calculate the percentage of time saved per task, per customer. Not impressions. Not engagement. Actual labor hours, replaced by machine intelligence.

The Engineer Who Learned to Count Everything

Ofer Feldman graduated from Technion - Israel Institute of Technology in 2001 with a BS in Computer Engineering. Technion is the kind of institution that produces disproportionate numbers of founders - it is Israel's engineering nerve center, and its alumni have shaped a significant portion of Silicon Valley's technical leadership over the past three decades.

His early career traced a deliberate arc through complexity. After military service in the Israel Defense Forces managing recruitment data systems, he moved into defense electronics at Elbit Systems, where software had to work the first time. From there he joined SAP Labs as a team lead and architect - a formative experience in enterprise software that would resurface years later in Stampli's obsessive focus on ERP integrations.

The decade before Stampli was a study in scale. At Tippers.com, he served as VP of R&D for roughly four years, growing the engineering team and learning what it means to build a product from garage-stage to something resembling a real operation. At McKesson, one of the largest healthcare companies in the United States, he managed agile teams building cardiology software - patient outcomes depending on code quality. High stakes. Methodical process. That discipline shows in how Stampli approaches accuracy.

Stampli: The Quiet Giant of AP Automation

Stampli was built on a specific bet: that accounts payable was broken in ways that no one in finance wanted to admit out loud. Invoices got lost. Approvers didn't know what needed their attention. Finance teams spent hours on tasks that should take minutes. The ERP systems that theoretically managed all of this had been designed for accountants, not for the people doing the work.

Ofer and his brother Eyal - who serves as CEO - approached the problem from both sides. Eyal, the business-facing co-founder, brought customer intuition. Ofer built the machine. Together they created something unusual in fintech: a platform that finance teams actually want to use. The product sits on top of existing ERPs - it doesn't replace them. That decision to integrate rather than compete with the infrastructure already in place is why Stampli now connects with 70+ ERP systems, from SAP and Oracle to QuickBooks and Acumatica.

The AI layer is where Ofer's years of careful, unsexy work pays off. Billy the Bot doesn't just extract invoice data. It learns each customer's unique coding patterns, approval hierarchies, and vendor relationships. It routes exceptions, flags anomalies, detects potential fraud, and gets smarter with every interaction. The system improves per customer, per workflow, over time - which is what actually differentiates AI in production from AI in a pitch deck.

In 2024, Stampli introduced Cognitive AI at SuiteWorld - a system that combines large language models with structured business logic to handle complex PO matching scenarios that previously required human judgment. This is not the AI of chatbot demos. It is AI doing the specific, consequential work that finance teams care about: matching a three-way purchase order against a receipt and an invoice when the numbers don't perfectly align.

By 2025, Stampli had expanded beyond AP automation into full procure-to-pay - covering requisition through payment in a single platform. This represents the logical endpoint of what Ofer has been building for a decade: not just smarter invoice approvals, but an entire financial workflow from the moment a company decides to buy something to the moment the vendor gets paid.

$145.7M and the Vote of Confidence

The Series D tells the most interesting story. In October 2023 - a period when fintech funding was contracting sharply across the board - Stampli closed a $61 million round led by Blackstone. Blackstone is not a typical fintech VC. It is one of the largest alternative asset managers in the world, with deep operational familiarity with exactly the kind of finance complexity that Stampli addresses. When Blackstone leads your Series D, it signals something about enterprise credibility that no press release can manufacture.

Existing investors Insight Partners, SignalFire, Bloomberg Beta, and NextWorld Capital all participated - a clean signal from people who had already seen the numbers up close.

Measure Everything. Hype Nothing.

There is a particular way Ofer Feldman talks about AI that distinguishes him from most people building AI products in 2024. He measures it by actual time saved per task. Not model accuracy on benchmarks. Not number of integrations. How long did a specific task take before Stampli? How long does it take now? Calculate the percentage. Report the savings. Repeat.

This is the engineering mindset applied to AI product development: if you can't measure it in terms a finance director would recognize, it doesn't count. "We're not building AI for AI's sake" is the kind of sentence that sounds obvious until you've spent five minutes watching enterprise software vendors claim AI transformation while their products require a consultant to configure a basic approval workflow.

The structural choices at Stampli reflect this philosophy. The data warehouses, user workflows, and internal engineering processes are all organized around one goal: delivering AI that does progressively more work on behalf of customers. Not AI that impresses in demos. AI that saves measurable hours, reduces measurable errors, and catches measurable fraud.

The Long Path to $100B

Things Worth Knowing

Follow the Work