AI Teammates Officially Launch: Proactive Collaboration, Task Ownership, and Slack Integration

From Tool to Teammate: The Shift in AI's Role

In the past, we were used to treating AI as a tool that's always on call—ask a question, get an answer, ask again. But the newly launched "Teammates" feature attempts to change this one-way interaction model. According to the official description, Teammates will "work for you," proactively taking on tasks and leveraging your Computer tasks, skills, memory, and tools to push work forward.

The core of this shift is proactivity. It no longer waits for you to input instructions, but like a real colleague, understands context, calls on resources, and moves things forward. The official description emphasizes "work proactively," meaning AI is beginning to have the ability to initiate tasks and follow through continuously, rather than just responding passively.

The Underlying Support for Proactive Work: Memory, Skills, and Tools

The reason Teammates can be "proactive" is inseparable from the integration of several key capability modules:

• Computer tasks: Enabling AI to operate computer interfaces and complete cross-application actions
• Skills: Preset or custom capability sets that let AI handle work in specific domains
• Memory: Retaining context and historical information to avoid starting from scratch every time
• Tools: Connecting to external services to expand the boundaries of AI's actions

These capabilities are not new, but packaging them into a "teammate" role represents an upgrade in product thinking. Previously, these features were scattered across different entry points, and users had to combine them themselves; now Teammates unifies them into a single proactive working entity, lowering the barrier to use and making AI's behavior closer to that of a real collaborator.

Computer tasks refers to the ability to let AI complete tasks by simulating human operations on graphical interfaces, typically involving screen understanding, clicking, typing, and cross-application switching. Unlike the traditional approach of calling services through APIs, Computer tasks enables AI to handle legacy software or web applications that don't expose interfaces, thereby greatly expanding the boundaries of what can be automated. The technical challenges of this capability lie in identifying and locating interface elements, planning operation sequences, and verifying and correcting operation results.

AI Agent refers to an AI system that can perceive its environment, autonomously plan, and execute multi-step operations to achieve goals. Compared with traditional Q&A assistants, the core characteristics of an Agent include goal orientation, tool invocation, state memory, and self-reflection. The "proactive work" described by Teammates is a typical manifestation of the Agent paradigm: it needs to break down tasks, select tools, and drive execution without step-by-step instructions, adjusting strategy based on intermediate results.

Collaboration Between Teammates: AI Can Also Message Each Other

A noteworthy detail is that Teammates can not only work with you but also collaborate with other teammates. The official description explicitly mentions that "Teammates collaborate with other teammates," which means multiple AI roles can divide work, pass information, and complete tasks together.

Furthermore, Teammates can also send messages to colleagues in Slack. This pushes AI from a personal assistant into team collaboration scenarios—it's no longer just a tool for you alone, but can intervene in team communication flows and proactively reach out to relevant people. For teams already deeply using Slack, this integration means AI can take on some of the coordination and follow-up work.

Multi-Agent Collaboration refers to a model in which multiple AI agents work together to complete complex work through message passing, task division, and result aggregation. Common architectures include an orchestrator responsible for task decomposition and allocation, workers responsible for specific subtasks, and a shared context or memory pool for information synchronization. The advantages of this model include parallel processing, specialized division of labor, and improved reliability through cross-validation; the challenges include communication overhead, task consistency, and the propagation of errors between agents. Teammates' support for "collaborating with other teammates" is the product-level implementation of this architecture.

Availability and Positioning: Web Version for Max Subscribers

Currently, the feature is "Available now for Max subscribers on the web." This information reveals two signals: first, Teammates is positioned as an advanced feature, prioritizing users on higher-priced tiers; second, the web version comes first, indicating the product is still being rolled out gradually, with no mention yet of mobile or other platforms.

From a product strategy perspective, placing proactive AI teammates at the highest subscription tier both matches their computational resources and complexity and serves as a differentiating selling point. For users considering upgrading their subscription, Teammates' proactive collaboration capabilities may become a key decision factor.

What This Means: The AI Agent Race Enters the Collaboration Phase

The launch of Teammates reflects a new direction in AI product competition: from "stronger single responses" to "more sustained autonomous work." When AI can remember context, call tools, proactively push tasks forward, and even communicate with other AIs and human colleagues, it begins to encroach on the territory of traditional assistants and project management tools.

Of course, the increase in proactivity also brings new questions—permission boundaries, task reliability, and the accuracy of message sending all need to be tested in actual use. The information officially disclosed so far is limited, and how it actually performs remains to be seen after Max users try it on the web. But what is certain is that the concept of "AI teammate" has moved from demo to a usable product form.

One-line Summary

Teammates turns AI from a passive tool into a proactive, collaborative AI teammate that takes on tasks.

Paragraph Summary

The article introduces the newly launched "Teammates" feature, which attempts to transform AI from an always-on-call tool into a proactive "teammate." Teammates proactively takes on tasks and integrates capability modules such as Computer tasks, skills, memory, and tools to push work forward. It can not only work with users but also collaborate with other AI teammates and send messages to colleagues in Slack, thereby intervening in team communication flows. Currently, the feature is available on the web for Max subscribers and is positioned as a premium differentiating selling point. The article argues that this reflects AI product competition shifting from "stronger single responses" to "more sustained autonomous work," but also points out that new issues such as permission boundaries, task reliability, and message accuracy still need to be tested in practice.