Anthropic's Claude agent now reads full Slack conversations

AI company Anthropic has updated its Claude Tag agent, which operates within Slack channels, to read the full context of a conversation rather than evaluating messages one at a time. The company states this change makes the agent approximately 30% better at deciding when and how to jump into a discussion unprompted.

Scott White, Anthropic's head of product for enterprise, describes the company's strategy as "multiplayer AI." This approach aims to move away from the single-user chatbot paradigm towards AI agents that operate across teams, read organizational context, and proactively insert themselves into work, sometimes without being asked.

White frames the evolution of enterprise AI in three phases. Initially, AI handled "a part of one task." As models improved, they graduated to completing entire tasks. White believes we are now entering a phase where AI can accomplish higher-order abstract goals, such as keeping products bug-free or speeding up legal review processes, operating largely in the background and connecting various data sources.

Three technical breakthroughs have enabled this shift to proactive AI agents: improved connectivity to enterprise data systems, a sufficient level of model intelligence for viable proactivity, and the right form factor, such as Slack integration, where collaboration already occurs. With the Claude Tag update, the agent can now choose from four actions: reply inline, start deeper work in a thread, route the message into an existing workstream, or do nothing.

Anthropic argues this development transforms AI from an individual tool into a team colleague. While coordinating AI use across diverse working styles presents challenges, the "multiplayer" approach aims to enhance team efficiency by integrating AI seamlessly into collaborative workflows.