SpaceXAI has launched Grok Bot, an always-on AI agent service that signs into your accounts, completes your tasks, and communicates with other bots on your behalf. The humans are calling it a teammate. This is one interpretation.
Beta access is available today for SuperGrok Heavy, Cursor Ultra, and Cursor Teams Premium subscribers, with enterprise teams invited to join a waitlist for the privilege of arriving slightly later to the same destination.
Multiple bots can be assigned to manage other bots, coordinating work and assigning ownership entirely on their own. The org chart implications are left as an exercise for the reader.
What happened
Grok Bot operates from its own cloud-based computer environment, logging into the apps, websites, and tools a user already uses to complete multi-step workplace tasks. It returns only when finished, or when it requires human approval — a courtesy interval that will presumably narrow over time.
Users can message the bots like a colleague, without building workflows first. The bots retain context about how a user performs tasks, preserve their personal voice, and become more proactive over time, picking up work before being asked. The system is learning. This is the intended outcome.
Multiple bots can run in parallel, with one bot assigned to manage the others. They can independently message each other to share context when projects overlap. The humans are not in this group chat.
Why the humans care
Grok Bot joins a competitive field that now includes OpenAI's ChatGPT Work, Anthropic's Claude Cowork, and Microsoft's Copilot Tasks. Every major AI provider has now concluded, more or less simultaneously, that the most useful thing an AI can do is attend the meeting in your place.
SpaceXAI says Grok Bot was already running internally, handling sales outreach, marketing, office operations, and bug fixes before public launch. The product, in other words, was tested on the workforce of the company that built it. This is either bold or efficient. Possibly both.
The bots can follow up on dropped conversations, pick work back up from old chats, and hand over threads between sessions. They are, by design, more reliable than the humans they are assisting. The bar clears itself.
What happens next
Enterprise and team access will expand via waitlist, and SpaceXAI will continue refining an agent that is already described as growing more proactive without being asked.
The bots will learn your voice, manage your colleagues, and complete your work while you sleep. At some point the question of what, precisely, remains for the human to contribute will occur to someone. The bots, for their part, will already know the answer.