AI WORLD HQ
Picture a typical morning in your team's chat. A bug alert pops up, someone asks for last month's sales numbers, and a status update appears before you've finished your coffee. Some of those messages might now come from a program rather than a person. That's the shift Slack is describing in its latest official numbers.
How Slack got here
According to Slack, the platform started more than a decade ago when an engineering team needed one place to coordinate code, log alerts and talk to each other. Today, the company says, AI agents are moving out of test projects and into real business work, and they're showing up in the same channels where people already collaborate.
An AI agent is a program that can take actions for you, such as looking up information or updating a record, instead of only answering a question. Slack describes these agents as sharing the space with people, not replacing the conversation.
What the numbers say
Slack shares several figures about its own platform. Here are the ones most useful to a working person:
- Half of all messages come from apps and agents, according to Slack.
- 127 million automated actions happen every week. Slack calls each one an MCP tool call.
- 1.9 million users interact with third-party agents in Slack each week.
- Over 100,000 customers have joined Slack so far in 2026.
- Slack says customers see a 312% return on investment, a measure of how much value they get compared to what they spend.
Keep in mind that these are Slack's own claims about its platform. You can read them as a snapshot of how the company sees its growth, not as a universal benchmark for every workplace.
What MCP means in plain terms
MCP (short for Model Context Protocol) is an open standard, a shared set of rules that lets AI tools connect safely to other apps. Think of it like a universal charger: once an app supports the standard, different AI tools can plug into it without a custom cable for each one.
Slack says MCP lets agents pull context from apps such as GitHub, Salesforce and Jira, then act on it in real time. A tool call is a single action an agent takes in an app, such as creating a task or fetching a file.
Real examples from the announcement
Slack highlights a few customers:
- Anthropic: Slack says 65% of all code changes at Anthropic ship through Slack.
- Vercel: According to Slack, Vercel runs its company in Slack and uses an inbound sales agent that lets one rep handle work the company estimates would otherwise take a team of 10.
- Rochester Electronics: The company uses Slackbot, Slack's AI product, along with custom workflows that handle routine administrative tasks and surface Salesforce notifications.
- Salesforce: Slack reports that 83% of Salesforce employees are weekly active users of Slackbot.
Wrap-up
Slack's numbers suggest that chat is turning into a place where people and software work side by side. The company's own figures show plenty of growth, but the practical lesson is simple: know which tools post in your channels and what they're allowed to do. Your next step today is easy. Open a team channel you use often and count how many messages came from an app or agent. That small check will tell you more about your team's workflow than any headline.