Salesforce quietly rolled out Anthropic’s Claude Tag inside Slack, even as it doubles down on its own Slackbot and Agentforce AI services. Claude Tag lets users type @Claude in any channel to break tasks into staged actions, follow up on unfinished threads, and surface relevant context from across workspaces. Unlike other bots, it learns from ongoing conversations, building institutional memory instead of starting fresh each time. That overlap puts Salesforce employees on edge. The company spent $27.7 billion on Slack in 2021 and has since built more than 30 AI features—meeting transcription, desktop-activity monitoring, CRM shortcuts—all running on Anthropic’s Claude model. Now customers face three competing helpers in one workspace: the legacy Slackbot, the enterprise-tailored Agentforce agents (which hit $800 million in ARR this quarter), and Claude Tag. Too many choices could slow corporate adoption. Money partly explains the move. Salesforce expects to buy $300 million worth of Anthropic compute this year and owns about 1% of the startup, now valued at $380 billion. By baking in a third-party agent that lives entirely within Salesforce’s security boundary, CEO Marc Benioff calls Slack “the interface to AI” and signals it’s model-agnostic. But the strategy blurs lines between in-house tools and partner offerings—raising internal questions about which products deserve top billing. Meanwhile Anthropic eyes expansion into Teams, email, and project-management apps, building an AI layer that could thrive without Slack at all.