Jeffrey Lai spent nearly a decade inside Apple, helping build the Chinese language version of Siri and leading other on-device AI initiatives. Now, he's betting that the next evolution of personal computing won't just respond to your commands—it'll watch you work, learn your habits, and eventually do the tedious stuff for you.
His startup, IrisGo, closed a $2.8 million seed round led by Andrew Ng's AI Fund this past January, valuing the Mountain View venture at $21 million. That figure might be conservative: IrisGo's own website claims $3.1 million raised from AI Fund, Pegatron, and other partners, hinting the round grew beyond its initial close. (The company declined to clarify the discrepancy.)
The pitch is deceptively simple. IrisGo builds software that runs locally on AI-equipped PCs—the new class of laptops and desktops with neural processing units baked into the chips. Show it a task once, clicking through the steps, and IrisGo converts that behavior into an automation. Do it again tomorrow, or next week, and the software quietly handles it. No cloud dependency. No uploading your workflow data to some distant server farm.
"Watch & Learn," they call it. Whether enterprises hungry for productivity gains will trust a startup to automate their knowledge work—well, that's the wager.
Apple DNA, Venture Studio Roots
Lai didn't leap straight from Cupertino to founding IrisGo. He and his co-founders, Arnold Yip (now CTO) and Yichen "Lman" Chu (COO), spent time as Founders-in-Residence at AI Fund, Ng's venture studio backed by north of $370 million from Sequoia, NEA, and SoftBank, among others. The incubator model gave them runway to tinker before spinning out as an independent company in 2025.
AI Fund's portfolio tilts heavily toward enterprise-facing AI tools, and IrisGo fits the mold: practical, incremental automation for businesses already spending heavily on new hardware. It's not flashy chatbot territory. It's the quiet labor of making computers a bit more... anticipatory.
Ng, the former Google Brain co-founder and Coursera co-creator, has built a reputation for backing applied AI rather than moonshots. IrisGo is very much in that tradition—grounded in the reality that millions of office workers still spend hours each week copy-pasting data between applications or running the same Excel macros.
Privacy-First, Chipmaker-Friendly
IrisGo's architecture leans on Intel's Core Ultra processors and the OpenVINO toolkit, running inference directly on the device rather than farming it out to the cloud. That's partly a privacy play—keeping sensitive workflows local—and partly a technical bet that the latest generation of PC chips can handle the computational load.
Intel has already published partner briefs showcasing IrisGo's "Digital Butler" software, the first appearing late last year. Acer signed on as a launch partner, and IrisGo is actively pursuing pre-installation deals that would put its software on machines fresh from the factory. In March, the startup joined NVIDIA's Inception program, adding another semiconductor giant to its ecosystem relationships. (Pegatron Ventures, the investment arm of the Taiwanese electronics manufacturer, also participated in the round.)
The company has woven in OpenClaw, an open-source automation framework that lets users share and run community-built workflows. It's a nod to the reality that no six-person startup—IrisGo's current headcount—can anticipate every use case. Better to let power users extend the platform themselves.
The AI PC Gold Rush (Maybe)
IrisGo's entire thesis rests on one assumption: that AI PCs will become ubiquitous, and soon. Gartner projected in late 2024 that AI-capable machines would represent 43 percent of all PC shipments by the end of 2025, with large enterprises expected to standardize on them by the following year. IDC analysts forecast AI PCs will dominate more than 90 percent of commercial shipments by 2028.
Perhaps. The PC industry has a long history of "next big things" that took longer to catch fire than vendors hoped. Remember ultrabooks? Or convertible tablets? But this time feels different, if only because Microsoft, Intel, AMD, and Qualcomm are all pushing hard in the same direction. Copilot+ PCs are Microsoft's flagship, but there's room for third-party software to differentiate beyond what Redmond bundles in.
IrisGo made the rounds at CES earlier this year, demoing its automation chops to OEMs looking for ways to stand out. The company is working from Mountain View now—not Palo Alto, despite the funding announcement—as it eyes broader manufacturer partnerships and an eventual Series A.
Whether enterprises will embrace workflow automation that literally watches over their employees' shoulders is another question entirely. Privacy concerns, IT security reviews, change management headaches—the path from seed funding to widespread adoption is littered with startups that had compelling demos but couldn't crack the go-to-market puzzle.
Still, Lai and his team have pedigree, patient capital, and chipmaker backing. In the race to make AI PCs actually useful rather than just buzzword-compliant, that's not a bad starting position.
