Local Vision Agent IrisGo Automates Desktop Workflows via NPU

On May 20, 2026, IrisGo emerged from stealth with a $2.8 million seed round and a new ambient desktop companion for macOS and Windows. Backed by Andrew Ng’s AI Fund and Nvidia, the startup replaces manual workflow scripting with an agent that learns by observing screen activity. The software targets the white-collar productivity gap by automating repetitive tasks that standard large language models cannot execute independently.

Local Vision and Hardware Integration

IrisGo relies on computer vision and on-device processing to monitor user behavior. The system watches a user perform a task once, such as ordering a specific item or processing an invoice, and can subsequently execute that task autonomously. To handle the compute requirements without compromising privacy, the application prioritizes local inference.

The software is optimized for the latest generation of AI PCs. It specifically leverages the NPU and GPU on Intel Core Ultra Series 3 (Panther Lake) processors to run AI inference locally. Complex tasks that exceed local capacity are routed through a hybrid cloud setup with end-to-end encryption, requiring explicit user authorization before execution.

Pre-Built Workflows and Execution

Instead of waiting for conversational prompts, IrisGo operates as background infrastructure. It senses context and coordinates actions across multiple third-party applications without relying on brittle API connectors. The beta release includes a predefined library of skills.

Capability	Function
Data Entry	Invoice processing and report generation
Communications	Email drafting and inbox management
Analysis	Automated document summarization
Development	Coding assistance similar to OpenAI Codex

At $20 per month for its premium tier, the company positions itself against Microsoft Copilot. While Copilot functions primarily as a chat-based assistant, IrisGo focuses on end-to-end task execution, behaving more like multi-agent systems operating across disparate local applications.

Funding and Distribution Strategy

The $2.8 million seed round values the company at approximately $21 million. Participation included Google for Startups and Pegatron Ventures, alongside Nvidia via its Inception program. Co-founder Jeffrey Lai previously contributed to the Chinese localization of Apple Siri, providing the technical foundation for the startup’s desktop integration.

To bypass the challenge of user acquisition, IrisGo secured a distribution partnership with Acer to explore preinstalling the software on new hardware. The company is actively negotiating similar deals across the PC ecosystem to establish direct hardware integration for AI agents at the operating system level.

If you manage corporate IT environments, hardware-bound agentic software changes your procurement strategy. Applications that rely on local NPUs require you to evaluate your endpoint hardware refresh cycles directly against the software automation capabilities your workforce needs.