From Chatbots to Ambient AI Agents
A new wave of ambient AI agents is reshaping how knowledge workers deal with email, calendars, and documents. Instead of waiting for prompts in a chat window, these assistants run in the background, watching how work gets done and quietly taking over repetitive tasks. The shift is from tool-based AI, where users must translate messy work into text prompts, to agent-based AI that lives alongside everyday apps and systems. These agents draft replies, move information between documents, and keep calendars in sync without constant user direction. Crucially, they aim to anticipate needs—suggesting actions or preparing content before you realize you need it. This ambient model is designed to make AI feel less like a destination you visit and more like an invisible layer of workspace automation that continually trims the busywork from your day.
IrisGo: A Desktop Companion That Learns Your Workflow
IrisGo is one of the clearest examples of this ambient approach. Rather than acting as another chatbot, it sits close to the operating system on AI PCs, observing how you navigate email, spreadsheets, documents, browsers, and internal tools. Over time, it learns patterns—like drafting emails from reports, pulling figures into slide decks, or repeating multi-step routines that usually require several tabs and many clicks. The assistant uses system accessibility features on Windows to interact with different applications and automate actions across the desktop, while emphasizing local processing and on-device learning to keep personal files and workflow context on the machine. That privacy-centric design is core to its pitch. Backing from Andrew Ng’s AI Fund and a reported seed round of USD 2.8 million (approx. RM12.9 million), along with an Acer distribution partnership, suggests major players see ambient desktop automation as a key frontier.
Google’s Gemini Spark: A 24/7 AI Email and Calendar Assistant
Inside Google Workspace, Gemini Spark pushes the same idea into the cloud office suite. Announced at Google I/O, Spark is described as a 24/7 personal AI agent that does more than answer questions. It can send emails, add calendar events, and complete tasks across Workspace apps, running long processes in the background without constant supervision. Google positions it as an AI email assistant and scheduling aide that takes action on your behalf but still asks permission before performing high-stakes tasks. Spark is arriving in preview for Workspace business customers through the Gemini app, tying into Gmail, Calendar, and other services. Alongside it, Google is rolling out conversational, voice-driven features like Gmail Live and Docs Live, further blurring the line between active prompting and ambient support as the system pulls context from Gmail, Drive, and the web to organize information automatically.
What Changes When AI Works in the Background
As ambient AI agents mature, the workday may feel less like managing apps and more like supervising a quiet digital colleague. Routine background AI tasks could include triaging inboxes overnight, pre-writing responses, pre-filling documents with the latest numbers, and assembling meeting notes before you sit down. But the upside comes with new expectations around privacy and control. A desktop assistant like IrisGo that learns from local behavior must be transparent about what it observes and when it acts, or it risks feeling like surveillance rather than help. Similarly, Workspace automation via Gemini Spark must clearly signal actions and ask consent where it matters. Trust will likely hinge on small details: clear logs of what the agent did, easy ways to override or correct it, and consistent reliability even when apps change. The promise is compelling—less friction, more focus—but the margin for error is thin.