IrisGo raises $2.8M to automate your desktop work

IrisGo automates – Backed by Andrew Ng’s AI Fund, IrisGo is betting on “proactive” AI agents that learn your PC routines and repeat them with limited or no prompting. The startup just rolled out beta apps for macOS and Windows and is pursuing deals to preinstall its desktop comp

A coffee order can be a small thing. But the demo IrisGo showed makes it feel like a rehearsal for something bigger.

On a PC screen. the system recorded the steps needed to place an order online: selecting a latte at Philz Coffee. filling in credit card information. and hitting purchase. Then the founder, Jeffrey Lai, asked Iris to do it again. The agent complied—without being walked through every step a second time.

IrisGo’s pitch is that this isn’t about coffee. It’s about turning one-off instructions into learned routines, then letting an AI desktop companion automate the repetitive parts of knowledge work—before you have to type out the same directions again and again.

The company is building a desktop companion for PCs that can learn about a user’s daily workflows and automate them with limited to no human prompting. IrisGo closed a $2.8 million seed round earlier this year, led by Andrew Ng’s AI Fund.

Iris was co-founded by Jeffrey Lai, a former Apple engineer who helped to build the Chinese language version of Siri, an automated assistant. The name “Iris” is also presented as a nod to that history—Siri spelled backwards.

To teach the system, the idea is straightforward: show Iris how to do something once. After that, it remembers the process for future automated use, aiming to remove the need for repeat instructions.

In day-to-day business settings. the targets are less romantic than ordering coffee but more constant: Iris comes with a built-in “skills” library that includes tasks such as email drafting. invoice processing. report building. document summarization. and other automated workflows. As the user works. the app also learns from desktop behavior and automatically adds those tasks to its potential list of action items.

There’s also a coding assistant included with Iris. The concept is positioned as similar to OpenAI’s Codex or Anthropic’s Claude Code, designed to help developers as they go about their work.

Lai frames the audience in plain terms. His target audience is knowledge workers at white-collar companies, where repetitive tasks pile up. He argues that even with today’s frontier models, AI-assisted office work can still feel manual. The goal. he said. is to shift toward a more autonomous workflow—where humans stay focused on high-level conceptual work while agentic systems handle the clerical steps in the background.

Privacy is part of the sales pitch too. Iris is designed to process a lot of data on-device. which Lai says gives stronger privacy protections than applications that rely heavily on cloud computation. The architecture is described as hybrid: larger, more complex tasks ultimately go through the cloud. But the company promises that cloud processing “only occurs when explicitly authorized by the user and uses end-to-end encryption.”.

That privacy promise lands alongside another strategy: credibility by association. Andrew Ng. a co-founder of the formative deep learning research team Google Brain. backed IrisGo—helping lead the seed round through his AI Fund. Lai said he set up a meeting with Ng through a shared connection: both are alumni from Carnegie Mellon University. During that meeting, Lai and his co-founder demoed Iris, and the AI Fund later led the seed financing. Nvidia and Google have also backed the company.

IrisGo is now moving from concept to installation. The startup recently launched beta versions of its macOS and Windows apps. It is also pursuing deals with laptop companies to preinstall the app on new devices. The company already struck such a deal with Acer. and Lai said the hope is to land similar deals with other device makers.

The story here is about timing as much as technology. IrisGo isn’t trying to replace the whole computer experience; it’s trying to take over the parts of it that already feel like busywork. If the system works as shown—learning a workflow after a single guided run—then “proactive” isn’t just a buzzword. It becomes something you notice every day, in the quiet moment when the tedious step is already done.

IrisGo Andrew Ng AI Fund proactive AI AI agents desktop automation macOS beta Windows beta Acer preinstall on-device AI cloud authorization end-to-end encryption coding assistant