In 2011, Dominic Lewis and two collaborators watched the SaaS explosion from a warehouse in South England and asked a question nobody else was asking loudly enough: who connects all of this?
The answer, in their reading of it, was nobody - or at least not at the speed the market demanded. Cloud applications were multiplying. APIs were proliferating. Data sat in silos that cost engineering teams months to bridge. Lewis, Rich Waldron, and Alistair Russell had been running a web agency together, and they'd seen the friction up close. So they stopped taking client work and started building Tray.io.
The early days read like a case study in resource-constrained ingenuity. The team worked from a shared warehouse space, selling Wellington boots on eBay to keep the lights on. When they finally decided to move to San Francisco - because that was where the enterprise market lived - they borrowed everything to make the flight and withdrew cash on credit cards to secure housing when they landed. There was no safety net. There was a product to build.
They had no obvious way into Silicon Valley's VC community, so they used their own automation tools to find and book meetings with investors. The irony was intentional: the platform they were selling, they were already using.
The Revenue Side of the Machine
Lewis has always occupied the commercial half of the Tray.ai equation. While Waldron leads strategy and Russell anchors engineering, Lewis drives the revenue function - building the go-to-market motion for a product that sits in the least glamorous but most essential layer of enterprise infrastructure.
Integration platforms don't get keynote slots the way consumer apps do. They don't trend on social. What they do is make everything else work. Cisco routes data through it. FICO builds decision workflows on it. Eventbrite automates its ticket operations through it. Segment uses it to move data. The list of companies whose operations quietly depend on Tray.ai reads like a Who's Who of the SaaS era - and selling into that market requires a particular kind of credibility.
Before Tray.ai, Lewis co-founded Plan.nr and worked at iFind Media. He studied at the University of Bath and attended Peter Symonds College before that - one of England's largest sixth-form colleges - graduating around 2004. The path from Winchester to San Francisco's enterprise tech scene is not an obvious one. But the scrappiness instilled in those early years appears to have been a feature, not a bug.
"Smart, scrappy and thinking about software integration the right way."
- True Ventures, describing the Tray.ai founding team at investmentFrom iPaaS to Agentic AI
The company Lewis helped build has shifted beneath him - and he's helped shift it. Tray.ai began as a visual workflow builder: a low-code platform for connecting cloud services without writing reams of custom code. It was a strong product for the 2015-2020 era of digital transformation. Then the AI era arrived, and the playbook changed.
Tray.ai's pivot to agentic AI orchestration isn't a brand refresh. The company rebranded from tray.io to Tray.ai to signal a fundamental repositioning: the platform isn't just connecting systems anymore. It's orchestrating AI agents that reason across those connections. The Merlin Agent Builder, launched in late 2024, lets enterprises build, govern, and scale AI agents in production. The Vanti acquisition in March 2025 - an Israeli AI startup specializing in knowledge modeling, named in ten Gartner Hype Cycles - added autonomous machine learning capabilities to the core.
That acquisition is worth sitting with. Vanti's technology allows AI agents to not just execute workflows, but to understand the business context behind them: to adapt, learn, and optimize in real time without human intervention at every step. In the architecture Lewis and his co-founders are building toward, the integration layer and the intelligence layer become one.
The Numbers That Define Scale
Tray.ai processes more than 150 billion workloads every year. That figure deserves a moment of comprehension: 150 billion automated tasks, running across 700-plus enterprise integrations, on a platform that has maintained 100% uptime. The revenue function Lewis manages sits atop infrastructure at genuine scale.
The company has raised $149.1 million in total, with the most recent round a $40 million Series C extension in September 2022. The round brought in Canada Pension Plan Investment Board alongside existing investors True Ventures, GGV Capital, Spark Capital, Meritech Capital Partners, and StepStone Group - a cap table that signals institutional confidence in the long game Tray.ai is playing.
Annual revenue sits around $16 million against that investment base and 150 employees. This is a company still in investment mode - building the infrastructure layer it believes will be foundational to the agentic enterprise. Lewis's job is to make the market believe it too.
What Thirteen Years Buys You
Staying power is underrated in startup culture. The dominant narrative rewards the exit, the pivot, the next act. Lewis has spent thirteen-plus years on a single thesis: that the connective tissue of enterprise software matters, and that building it right requires patience.
That's not stubbornness. The product has evolved radically from its origins as a general-purpose web automation tool. But the core conviction - that enterprises need a reliable, flexible, governed integration layer - has held. And as AI agents begin proliferating across enterprise stacks, the need for exactly that kind of infrastructure becomes more acute, not less.
The companies building AI agents need them to talk to Salesforce, to Workday, to SAP, to hundreds of legacy systems. They need governance so regulated industries can deploy them. They need security controls so data doesn't leak. They need audit trails so compliance officers can sleep. That's the platform Lewis has been building, in one form or another, since 2011.
When the founding team moved to San Francisco with no VC relationships and no contacts, they used Tray.io itself to automate outreach, find investor email patterns, and book meetings. The pitch deck sold a product they were already using as the tool that built the pitch.
The Architecture of the Agentic Era
The enterprise AI story of the mid-2020s is, at its core, a systems integration story. Every large language model that a Fortune 500 company wants to deploy needs to connect to live systems - CRMs, ERPs, data warehouses, ticketing platforms. It needs to read from them, write to them, trigger actions across them. It needs to do this at scale, with security controls, under governance frameworks that legal and compliance will approve.
Tray.ai's bet - and Lewis's commercial argument - is that you can't build reliable AI agents without a reliable integration foundation. The companies that tried to bolt LLMs onto brittle point-to-point integrations are discovering that in real time. The Merlin Agent Builder and the Vanti acquisition are both answers to the same question: how do you make AI agents that actually work in production, across real enterprise systems, at the scale enterprises demand?
Lewis has been selling into that market for years. The vocabulary has changed; the problem hasn't. Enterprises need their systems to talk to each other. They always have. The difference now is that AI agents are part of the conversation.