A new startup backed by prominent artificial intelligence figure Andrew Ng is developing software that observes a user’s computer screen to learn how to automate routine tasks. The company, IrisGo, is positioning its product as an autonomous desktop assistant.
The startup’s software, currently called Iris, was initially described by its co-founder as an “AI butler.” The program works by monitoring activity on a user’s desktop interface. Over time, it learns to replicate specific workflows and tasks without requiring manual programming or script writing.
How the AI assistant learns
According to the co-founder, Iris does not rely on predefined macros or static rule sets. Instead, it watches the visual user interface and builds a model of the user’s behavior. Once the system has learned a recurring pattern, it can perform that action autonomously in the future. This approach is designed to reduce the amount of time users spend on repetitive digital chores.
The technology falls within a broader category of AI applications often referred to as “agentic” systems. These are programs that can act on behalf of a user to complete specific goals. IrisGo is attempting to bring this concept to the everyday desktop environment, focusing on accessibility for a general audience.
Investor background and market context
Andrew Ng is a well known figure in the artificial intelligence industry. He co-founded Google Brain and served as chief scientist at Baidu. His involvement with IrisGo provides the startup with significant credibility in the competitive AI software market. Ng has previously invested in other ventures that aim to democratize AI tools.
The startup faces a landscape in which other companies are also developing similar agentic desktop assistants. These range from experimental projects in academic labs to products from major technology firms. IrisGo’s differentiator is its emphasis on learning directly from user behavior rather than requiring explicit training data or commands.
Implications for user workflow
For workers who rely on repetitive computer interactions, such as data entry, file management, or standardized report generation, an assistant that learns visually could reduce the time spent on these tasks. The software aims to operate without needing deep integration into existing applications.
The co-founder confirmed that the system is designed for privacy. While it watches the desktop, the developers have stated that data processing is handled locally on the user’s machine. This approach is intended to address common concerns regarding cloud based monitoring and data leakage.
Next steps for the project
IrisGo is currently in an early stage of development. The company has not announced a formal public release date. The immediate focus is on refining the learning algorithms and expanding the range of tasks the software can reliably handle. As the startup moves forward, it will likely need to demonstrate both the accuracy of its automation and the strength of its security model to win over a broader user base.
Source: GeekWire
