Salesforce employees 'confused' as company promotes Anthropic's Claude inside Slack, the very platform it paid $27.7 billion for; says: We can't tell where…
Salesforce is doing something odd right now. It's cheering on a product that competes with its own. When Anthropic rolled out Claude Tag last month—an AI agent that lives permanently inside Slack channels—Salesforce promoted it across social media. The problem? Salesforce owns Slack. And according to The Information, some employees inside the company are openly confused about why their employer is boosting a rival that eats into both Slackbot and Agentforce , the two AI products Salesforce has poured money and years into building.
The awkwardness isn't hard to spot. Slack customers can now pick between Slackbot, Agentforce agents, and Claude Tag—three AI assistants, one workspace, all fighting for the same job.
Why Salesforce's Claude Tag promotion feels so strange to its own employeesSalesforce paid $27.7 billion for Slack back in 2021, then spent years turning Slackbot into something smarter. In March, the company announced more than 30 new AI features for Slackbot—meeting transcription across Zoom and Google Meet, desktop activity monitoring, task execution through third-party tools via MCP, and even a lightweight CRM baked into small-business channels. It was the biggest overhaul since the acquisition, and Slack pitched Slackbot as an "agentic operating system"—a single surface for every AI agent and app a worker touches. Slackbot became the fastest-adopted product in Salesforce's 27-year history, with some customer teams reporting savings of up to 90 minutes a day.
Here's the twist: every one of those features runs on Anthropic's Claude. Anthropic's technology powers the reasoning layer while Slack handles the "context engineering"—deciding which messages, files, and channels get fed into the model. So Salesforce isn't just tolerating Anthropic. It's built on top of it.
Claude Tag now offers a parallel version of that same experience. Type @Claude in any channel and it takes your task, splits it into stages, and works through them where everyone can watch. There's also an ambient mode that jumps into conversations on its own—flagging updates, pulling context from other channels, chasing down threads people forgot about. Unlike a lot of assistants, it holds onto institutional knowledge instead of starting cold every time.
The money reasons behind Salesforce backing Anthropic's rival AI toolFor all the internal head-scratching, Salesforce has real incentives to play nice. The company expects to spend roughly $300 million on Anthropic tokens this year, and it holds about a 1% stake in Anthropic, now valued at $380 billion. CEO Marc Benioff has called Slack "the interface to AI" and framed the platform as model-agnostic—happy to host outside AI agents next to its own. Anthropic is also the first LLM provider fully sealed inside the Salesforce trust boundary, so data never leaves the ecosystem or feeds model training.
The awkwardness isn't hard to spot. Slack customers can now pick between Slackbot, Agentforce agents, and Claude Tag—three AI assistants, one workspace, all fighting for the same job.
Why Salesforce's Claude Tag promotion feels so strange to its own employeesSalesforce paid $27.7 billion for Slack back in 2021, then spent years turning Slackbot into something smarter. In March, the company announced more than 30 new AI features for Slackbot—meeting transcription across Zoom and Google Meet, desktop activity monitoring, task execution through third-party tools via MCP, and even a lightweight CRM baked into small-business channels. It was the biggest overhaul since the acquisition, and Slack pitched Slackbot as an "agentic operating system"—a single surface for every AI agent and app a worker touches. Slackbot became the fastest-adopted product in Salesforce's 27-year history, with some customer teams reporting savings of up to 90 minutes a day.
Here's the twist: every one of those features runs on Anthropic's Claude. Anthropic's technology powers the reasoning layer while Slack handles the "context engineering"—deciding which messages, files, and channels get fed into the model. So Salesforce isn't just tolerating Anthropic. It's built on top of it.
Claude Tag now offers a parallel version of that same experience. Type @Claude in any channel and it takes your task, splits it into stages, and works through them where everyone can watch. There's also an ambient mode that jumps into conversations on its own—flagging updates, pulling context from other channels, chasing down threads people forgot about. Unlike a lot of assistants, it holds onto institutional knowledge instead of starting cold every time.
The money reasons behind Salesforce backing Anthropic's rival AI toolFor all the internal head-scratching, Salesforce has real incentives to play nice. The company expects to spend roughly $300 million on Anthropic tokens this year, and it holds about a 1% stake in Anthropic, now valued at $380 billion. CEO Marc Benioff has called Slack "the interface to AI" and framed the platform as model-agnostic—happy to host outside AI agents next to its own. Anthropic is also the first LLM provider fully sealed inside the Salesforce trust boundary, so data never leaves the ecosystem or feeds model training.
Next Story