IrisGo, an Andrew Ng-backed startup, looks to become the AI-powered desktop companion you never knew you needed

Industry insiders say the next big thing in AI is “proactive” systems: agents that can anticipate a user’s needs — and meet them — even before the user knows what those needs are.

One startup looking to make headway in this area is IrisGo. The company, which closed a $2.8 million seed round led by Andrew Ng’s AI fund earlier this year, is building a desktop companion for computers that can learn about a user’s daily workflow and then automate them without any human prompting.

Iris was co-founded by Jeffrey Lai, a former Apple engineer who helped build the Chinese version of Siri, the company’s robotic assistant. (Somewhat disingenuously, Iris is Siri spelled backwards.)

The basic idea is simple: Show Iris how to do something once, and she’ll remember that process for future automated use — without having to repeat instructions.

During a conversation with TechCrunch, Lai did a demo showing how Iris can learn how to order coffee online. While watching, Iris recorded the steps she took to select a latte from Philz Coffee (a popular chain in the Bay Area), fill out the credit card information, and then hit the purchase button. Lai then asked Iris to repeat it herself; The agent dutifully complied.

Buying coffee, of course, is not the point. Instead, we hope the system will automate a whole host of business-related tasks. Iris comes with a library of built-in “skills” — things like email drafting, invoice processing, report generation, document summarization, and many other automated workflows that are ready to use. Meanwhile, Iris learns from the user’s desktop behavior and automatically adds those tasks to its list of potential action items.

The app also includes a programming assistant — similar in concept to OpenAI’s Codex or Anthropic’s Claude Code — designed to assist developers as they do their work.

“Our target audience is knowledge workers — white-collar companies. There are a lot of repetitive tasks that these workers do every day,” Lai said, noting that despite the high power of today’s frontier models, AI-assisted office work still feels incredibly manual and repetitive. The goal, he said, is to move away from that and toward a more autonomous workflow, where humans work on high-level conceptual work while agent systems take care of all the clerical work in the background.

A particularly attractive feature of Iris is that it’s designed to process a lot of data on the device, giving it stronger privacy protections than other apps that rely heavily on the cloud. Lai says the system is still a hybrid architecture — meaning that larger, more complex tasks are eventually processed through the cloud, though the company promises that cloud processing “only occurs when explicitly authorized by the user and uses end-to-end encryption.”

Part of Iris’s expansion strategy was to gain credibility by associating with prominent figures and organizations. This was helped by support from Ng, co-founder of the configural deep learning research team Google Brain. Lai was able to arrange a meeting with Ng through a common connection: both are Carnegie Mellon graduates. Lai and his co-founder pitched Iris during that meeting, and ING AI Fund eventually led the startup’s seed round. Nvidia and Google have also backed the company.

IrisGo recently launched beta versions of its macOS and Windows apps, and the company is also currently seeking deals with laptop companies to pre-install the app on new devices. It recently struck such a deal with Acer, and Lai said the hope is that the company can strike similar deals with other hardware makers soon.