Anthropic brings Claude inside Slack channels

Anthropic has kicked off a research preview of Claude Tag, an integration that turns its LLM into an always-on teammate within Slack. Already available to Claude Enterprise and Claude Team customers, the feature lets users mention @Claude in any channel to get insights, generate text, or delegate tasks. The announcement places the company among the providers of AI that is increasingly embedded in daily workflows, but it also reignites questions about data governance when the assistant lives entirely in the cloud.

A teammate that lives in chat

The idea behind Claude Tag is a persistent AI teammate, not confined to a single interaction in a separate chat window. With the @Claude syntax, the LLM participates in channel conversations, retains context, and can be summoned on demand, much like a human colleague. Anthropic presents the feature as an evolution of its existing integrations, aiming to reduce the friction that often keeps AI tools isolated and underused. For businesses, it means an assistant that doesn’t need to be found – it’s already embedded in the most widely used collaboration platform.

The convenience of always-on AI and its limits

An always-on assistant promises instant answers, fewer interruptions, and faster decision-making. However, persistence also means the model constantly processes channel messages, including sensitive ones. Anthropic has not yet disclosed specific details on data handling, but because the service operates in the cloud and interacts with Slack, content travels through servers outside the customer’s own infrastructure. For teams in regulated industries such as finance or healthcare, the lack of guaranteed data residency can make Claude Tag hard to adopt without contractual reviews or additional safeguards.

The data factor: cloud or on-premise?

The arrival of Claude Tag spotlights a growing trade-off for those evaluating LLMs in the enterprise: hand over infrastructure management in exchange for ready-to-use integrations, or retain full control with on-premise deployment. Self-hosted solutions built on open models, frameworks like vLLM or Ollama, and dedicated hardware now allow local inference, meeting GDPR and data residency requirements. However, those who choose on-premise lose the simplicity of an assistant natively attached to Slack, unless they develop custom connectors. Claude Tag thus becomes a litmus test for the enterprise market: the ease of a cloud service can win, but only if privacy concerns are addressed transparently.

Beyond Slack: the race for integrated assistants

Anthropic’s move is not isolated. Google with Gemini in Workspace, Microsoft with Copilot in Teams, and startups focused on AI agents are redefining collaboration. In this landscape, LLM builders are vying to expand their footprint across productivity platforms, creating an ecosystem where AI becomes a horizontal layer. It remains to be seen whether Anthropic’s Slack path will lead to more modular interfaces, perhaps with the option to choose on-premise inference endpoints – a step that could combine the best of both worlds and resolve the dilemma many companies now face.