What is an always-on agent?

An always-on agent is an AI agent that keeps working with no person in the conversation: it runs in the background, picks up work from standing goals, schedules, or events, acts on it, and comes back to you with a result or a request for approval.

OpenAI uses the term directly: its September 29, 2026 announcement calls dots “always-on agents” that run on their own cloud computer, connect to over 4,000 apps through plugins, and do read-only “proactive research” when you aren’t working with them. Google uses other words for the same shape. Its Gemini agent, announced at Gemini at Work 2026 in an October 9 post, can be assigned work, scheduled, or set to respond to events, and keeps running for hours or days after you close your laptop; a “coworker agent” version holds a persistent role with its own identity and email address. Google’s AQuA (Ambient Quality Agent), published on October 8 in the google/adk-recipes repository, is a narrower case: it runs beside a production agent and reviews a sample of up to 1,000 recent sessions on a schedule, after each deploy, or on demand.

Because nobody reads each result as it happens, dots ships an Activity View for following background work, and AQuA keeps a dashboard of the issues it has found. An agent that runs around the clock can be up the whole time and still wrong, and that only shows when someone looks.