One Company Now Has More AI Agents Than Human Employees | Ryan Gavin of Slack

What It Covers

Slack CMO Ryan Gavin explains how Slack is evolving into an agentic operating system for enterprises, where Slackbot now accesses structured and unstructured company data, orchestrates third-party and Salesforce AgentForce agents via MCP, and enables individual employees to direct multi-agent workflows through natural conversation.

Key Questions Answered

• •Agentic Orchestration via MCP: Slackbot functions as an MCP client, meaning any MCP-compatible agent — from OpenAI, Anthropic, Perplexity, or custom enterprise builds — can be invoked through a single conversational interface. Employees at-mention agents the same way they at-mention colleagues, eliminating the need to locate, learn, or switch between separate agent tools.
• •Unstructured Data as Competitive Moat: Slack's 6 billion weekly messages represent a largely untapped long-term memory layer. Slackbot now indexes this conversational history alongside structured CRM data, enabling context-rich outputs — such as full customer relationship summaries — that no standalone LLM can replicate without that internal organizational context.
• •Employee Productivity Benchmarks: MrBeast Industries employees report saving 20 hours per week — equivalent to six months annually — using Slackbot. Salesforce engineering teams report building product capabilities in two days using seven specialized agents (engineering, design, testing) that would previously have required a full separate company to deliver.
• •Reusable Skills as Personal Agent Building: Employees can create reusable Slackbot skills — essentially mini-agents with defined inputs, formats, and outputs — without coding. A marketing employee built a data scientist agent over one weekend, enabling 89–90% of cohort analysis to run autonomously, with the human data science team only validating final outputs rather than executing the full workflow.
• •Agent-to-Employee Ratio as a New Metric: At least one large, named AI company now operates with more deployed agents than human employees. Enterprises should treat agent headcount as a measurable workforce metric and use Slackbot's orchestration layer to manage agent discovery and task routing, preventing the same fragmentation problem that plagued enterprise SaaS sprawl.