IrisGo and the Rise of the Proactive AI Desktop Assistant

The product is designed as an "AI desktop buddy" for macOS and Windows, helping users manage repetitive knowledge‑work tasks across multiple applications instead of operating within a single chat window or tool.

Examples of tasks the assistant may perform include:

• Drafting or replying to emails
• Processing invoices
• Building reports
• Summarizing documents
• Ordering items or handling routine requests
• Assisting with coding tasks

These capabilities position IrisGo as a productivity assistant focused on automating workflows rather than just answering questions.

How the Proactive Assistant Learns Workflows

The central idea behind IrisGo is that the assistant learns from observing a user’s actions on their computer and converts repeated processes into reusable automation routines.

According to reports and company descriptions, the workflow typically works like this:

1. A user demonstrates a task—such as generating a report or processing an invoice.
2. IrisGo observes the steps taken across applications.
3. The system records those steps as a reusable "skill" or workflow.
4. In the future, the assistant can perform the same task automatically or with minimal prompting.

The approach resembles automation tools but aims to be more flexible. Instead of requiring scripts or rigid rule‑based macros, IrisGo uses AI to understand context across apps and workflows, allowing it to coordinate actions across documents, browsers, communication tools, and internal software.

This design aligns with a broader trend toward “proactive AI agents”—systems that anticipate needs and take action without constant prompts from the user.

On‑Device AI and Hybrid Privacy Design

A major part of IrisGo’s positioning is its focus on on‑device AI for personal computers, sometimes referred to as “AI PCs.”

In this model, much of the assistant’s processing happens directly on the user’s machine rather than entirely in the cloud. This approach can improve privacy because sensitive data stays local to the device whenever possible.

IrisGo’s documentation describes a hybrid architecture combining local models with cloud AI, where:

• Contextual data such as screen content, clipboard information, or voice input is processed locally whenever possible.
• Some AI tasks may still rely on cloud models when needed.
• Users must explicitly initiate certain contextual processing actions.

The company also notes that screen and audio data are not continuously recorded, and that captured data is typically stored locally unless a user chooses to send it to an AI model or enable cloud features.

However, public sources provide limited technical detail about the exact architecture, such as the precise division between local and cloud computation or the internal permission system governing automated actions.

Founders and Origins

IrisGo was co‑founded by Jeffrey Lai, a former Apple engineer who helped develop the Chinese‑language version of Siri.

The startup operates within the ecosystem of Andrew Ng’s AI Fund, which builds and backs early‑stage AI companies focused on practical applications of machine learning.

The project’s name reportedly references Siri in reverse (“Iris”), reflecting its ambition to build a next‑generation assistant focused on productivity and automation rather than simple voice commands.

Target Users

IrisGo appears aimed primarily at knowledge workers who spend large portions of their day managing digital workflows.

Typical target users include people who regularly:

• Handle large volumes of email
• Generate reports or presentations
• Process invoices or documents
• Conduct research or summarization
• Manage purchasing or administrative tasks
• Write or review code

Because the assistant operates across desktop apps, it is intended for workflows that span multiple tools rather than tasks confined to a single SaaS platform.

Distribution and AI‑PC Partnerships

IrisGo is developing desktop applications for both macOS and Windows, with early beta versions available for download.

The company has also explored distributing its assistant through partnerships with laptop manufacturers, allowing the software to ship preinstalled on certain computers.

Some reports mention potential collaborations with PC brands such as Acer, though publicly documented confirmation of large‑scale preinstallation agreements remains limited. As with many early‑stage startups, these partnerships appear to still be exploratory rather than finalized distribution deals.

What Remains Unclear

While the vision for IrisGo is well described, several details about the product remain unclear from available public information:

• The full technical architecture behind its automation system
• The precise balance between on‑device and cloud AI processing
• Supported third‑party applications and integrations
• Enterprise features or pricing plans
• The timeline for widespread commercial release

Until more technical documentation or large‑scale deployments appear, most descriptions of the assistant should be viewed as reported capabilities and demonstrations rather than fully verified production features.

Why IrisGo Matters

IrisGo represents a broader shift in AI development: moving from reactive chatbots toward agents that understand a user’s environment and proactively perform tasks.

If successful, this type of system could transform personal computing—turning desktop assistants into automation layers that watch how people work and quietly handle routine digital labor in the background.

For now, IrisGo is an early example of that idea, combining workflow learning, on‑device AI, and cross‑app automation into a prototype of what a truly proactive AI assistant might look like.