The Idea
In early 2025, Bar Winkler and Roey Lalazar looked at the enterprise AI boom and noticed something nobody was pricing in: almost all of it was being built in English, for English. Winkler had run global expansion at ironSource before founding Approve.com and selling it to Tipalti in 2021. Lalazar had built Kaps, an AI localisation platform. Between them they had spent years selling software into Europe, the Middle East and Latin America, and they knew what happened when a US-built product met a Portuguese call centre or a Hebrew support desk. It broke. Not technically, culturally. The demo worked. The deployment didn't.
Their read was that most of the world was sitting in what they called "AI deserts", regions where the technology existed but nobody had done the unglamorous work of making it land in the local language, under local regulation, with local norms. Everyone else was queuing up to sell to the Fortune 500. They decided to go everywhere else, and to put their own engineers in the room to make it work.
This was the beginning of Wonderful.
The Execution
- July 2025: Wonderful came out of stealth with a $34 million seed led by Index Ventures, with Bessemer and Vine Ventures joining. It was already powering hundreds of thousands of customer interactions for 15 enterprises across voice, chat and email.
- November 2025: Four months later, a $100 million Series A, again led by Index, with Insight Partners and IVP joining. Total raised hit $134 million. The market was crowded with AI agent startups and the open question was whether this was another GPT wrapper. Wonderful's answer was distribution: it was live in 30 countries inside a year, with local teams on the ground in each one.
- March 2026: Insight Partners led a $150 million Series B at a $2 billion valuation, 13 months after founding. Headquarters moved to Amsterdam. The plan was to go from 350 to 900 staff by year end.
- March 2026: The land-and-expand numbers were the real signal. Winkler said over 70% of enterprises starting with a single use case added more workflows within three months, because the integrations were already built.
- Mid 2026: The company outgrew its original product. Agents handling support tickets became a full AI operating system: managed workflows, productivity agents, AI-native applications that replace legacy software, and conversational agents, all sharing one governance layer.
- Mid 2026: The go-to-market copied a very specific playbook. Forward-deployed engineering pods embed with the customer, ship the first use case, then hand over the keys, a model that closely tracks Palantir's AIP rather than classic SaaS.
- 2 September 2026: Insight led a $550 million Series C at a $5 billion valuation, its second round leading in a row. Salesforce came in as a new investor alongside Index, IVP, Bessemer, Vine and 9Yards.
- Today: 650 employees across more than 35 markets, roughly half the team in Israel, and over $800 million raised in 20 months. The valuation went up 2.5x in six months.
The lesson?
Everyone else was fighting over the same English-speaking enterprise buyer. Winkler and Lalazar went to the markets where the demo didn't work, then sent their own engineers to sit with customers until it did. That second part is the bit most founders skip because it doesn't scale cleanly and it looks like consulting on a spreadsheet. It's also why 70% of their customers expanded within three months. The model was never the moat. Being in the room was.