Build a Scalable Agentic Work Operating System in Slack

Slack marked the first anniversary of its agentic-era platform by detailing native Model Context Protocol (MCP) support, including an MCP Server, a Real-Time Search API, and Slackbot as an MCP client, with more than 40 Marketplace partners such as Atlassian Rovo, Box, Canva, DocuSign, Figma, Hibob, Linear, Notion, Pagerduty and Zoom plugging in directly. Slack said the platform connects foundation models including Claude, ChatGPT and Gemini alongside specialists like Cursor, Devin, Perplexity and Agentforce, with Marketplace access to 6,000+ apps and agents, and cited Salesforce data that companies run more than 1,000 separate software applications and Gartner's forecast that global AI spending approaches $2.5 trillion this year. Slack framed the offering as a way for IT leaders to drive adoption and prove ROI on existing AI investments while keeping data unified and governance at the identity layer.

I spent the last week at Dreamforce in conversations with customers wrestling with the same problem: they’ve invested heavily in AI, but most of it still lives in a browser tab, disconnected from how their teams actually work. Companies are running more than 1,000 separate software applications https://www.salesforce.com/ap/blog/2025-connectivity-benchmark-report/ right now, and global AI spending is approaching $2.5 trillion https://www.gartner.com/en/newsroom/press-releases/2026-05-19-gartner-forecasts-worldwide-ai-spending-to-grow-47-percent-in-2026 this year. Most of that money is already spent, so the next challenge facing IT leaders is driving adoption and proving ROI on what they’ve already bought. Much of that value stays locked away because tools sit in isolation, outside the daily flow of work. That’s where Slack comes in. A year ago, we unveiled an updated Slack platform https://slack.com/features/agentic-platform built for the agentic era, where humans and agents work together in the same place conversations already happen. Over the last year, we’ve delivered a work operating system connecting your people, agents, and every platform you use, all in one place. In Slack, everyday conversations become the shared context powering your tech stack. This is your blueprint for maximizing existing software investments, protecting data boundaries, and building an agentic architecture you can stand behind. Build for the future with a flexible, open platform The enterprise AI landscape shifts weekly. Lock your infrastructure into one vendor or proprietary LLM, and you accumulate technical debt fast. CTOs need the freedom to swap models and tools on demand, without breaking downstream workflows. Built on an open, secure, model-agnostic philosophy, Slack lets you bring AI agents your team already relies on, connect your preferred models, or dive straight in with Slackbot. At the center of that fabric is native support for the Model Context Protocol MCP , the open industry standard connecting AI models to tools and enterprise data. With MCP in Slack you can: - Ground agents in your conversational data https://slack.com/blog/news/mcp-real-time-search-api-now-available : Generalist LLMs miss what your team decided last week. They weren’t in the huddle. Slack’s MCP Server and Real-Time Search API bridge that gap, enabling any AI provider to securely read and synthesize threads and decisions in real time. No data replication, and no guesswork. - Coordinate work through Slackbot https://slack.com/blog/news/slackbots-mcp-client : As an MCP Client, Slackbot turns plain-language requests into action, routing your ask to the right tool without you leaving the conversation. More than 40 Marketplace partners – including Atlassian Rovo, Box, Canva, DocuSign, Figma, Hibob, Linear, Notion, Pagerduty and Zoom – plug directly in. - Connect your stack with a rich app and agent ecosystem https://slack-ce.slack.com/marketplace : Deploy foundation models like Claude, ChatGPT, and Gemini alongside specialists like Cursor, Devin, Perplexity, and Agentforce, with direct Marketplace access to 6,000+ apps and agents. Model execution and interface are decoupled, so you can swap backend infrastructure or add specialized agents without retraining anyone. Data remains unified, governance lives at the identity layer, and your stack adapts as fast as the market does. The result is an AI platform that grows with your business instead of forcing you to rebuild it. Shift from single-player chats to multiplayer velocity Most AI tools are single-player: a browser tab where institutional knowledge evaporates the moment the session ends. While AI alone makes one person smarter, AI in the open, within a Slack channel, makes the entire company smarter. Slack transforms AI execution into a multiplayer experience. In open channels, agents work alongside teams, keeping their reasoning, outputs, and decisions visible, searchable, and actionable. More people in a channel means richer context and better outputs, each turn compounding organizational value. Work moves at the speed of the team, not the slowest handoff. In August, we extended multiplayer AI for software creation with Slack Code https://slack.com/features/code-channels . Cramming development work into standard channels creates noise, while letting developers retreat into private tabs loses shared context. Dedicated code channels solve both by giving teams and coding agents Claude, Devin, GitHub, Vercel a purpose-built space to plan, review, and ship together, complete with embedded canvases, code diffs, and live HTML previews. Picture a product manager spotting a bug report in a channel. Instead of filing a ticket and waiting on the next sprint, they tag a coding agent, which spins up a code channel, reviews the thread history, and proposes a fix. An engineer verifies it in a live preview and approves the deployment. Once merged, the channel archives itself, leaving a searchable audit log for the whole company. Slack becomes the one place where your stack, your teams, and your agents work as a single system. And everyone learns when AI operates in public. Make Slack the front door to Salesforce and enterprise data For decades, technology leaders have invested heavily in Salesforce, custom logic, customer records, and automated workflows. But that value stayed locked inside the CRM: teams in Product, Marketing, Engineering, and leadership rarely logged into Lightning Experience to find it. Work is conversational, and Salesforce Hosted MCP Servers bring Salesforce data natively into Slack https://www.salesforce.com/slack/new-mcp-servers-ai-data-in-slack/ . Every team can access, update, and act on CRM insights, Data Cloud pipelines, and Tableau metrics right where they already collaborate, with zero custom integration code. Here’s what you unlock: - Salesforce CRM: review, create, or update account details in-thread, no Lightning login required. - Data 360: non-CRM users get instant, real-time customer profiles unifying CRM, marketing, and external data. - Tableau Next: live KPIs and visual reports rendered directly into daily chat. - Custom Enterprise Actions: run Apex Actions, Lightning Flows, and Agnetforce agents from a simple text prompt. Extending CRM visibility doesn’t create new exposure. Every action still respects your existing permissions, field-level security, and sharing rules. Admins keep full control in the Slack Admin UI. Everything you've built with Salesforce — your data, your workflows, your agents — is now available in Slack the same way you'd talk to a teammate. — Rob Seaman, EVP and GM of Slack Give everyone the tools to build the agents they need Text-only chatbots that just talk about work don’t deliver ROI. Real productivity happens when AI takes action: resolving incidents, updating records, or triaging support tickets directly in the flow of work. That means teams closest to the problem need to build their own solutions, raising the classic IT dilemma: how do you give marketing, HR, and sales the tools to build agents without opening the floodgates to unvetted AI? Slack solves this with a single enterprise-grade development layer for both pro-code engineers and no-code builders – with security built in, not bolted on. - Developer Tools for Interactive Agents : http://slack.dev build production-grade, action-oriented agents in minutes with the Bolt SDK and Slack Developer Kit for Agents. Ship interactive Block Kit UIs and Work Object Embeds live records inside Slack threads. - No-Code Deployment for Business Teams https://slack.com/blog/news/add-to-slack : Slack’s “Add to Slack” button lets business teams deploy agents built on Vercel, OpenAI, and Lovable with one click – no engineering ticket. Permission scoping, SSO, and data boundaries apply automatically. Because technical and non-technical teams build agents that operate inside shared channels, execution happens out in the open. Agents surface their reasoning where the whole team can approve actions. That’s a safe way to scale autonomous AI with people in the loop. The strategic path forward for IT leaders The shift to an agentic enterprise isn’t about deploying different or more models; it’s about changing where work gets done. Agents stuck in private tabs create friction and hide the reasoning your teams need. Slack is where AI capability actually turns into work. It brings models, data, and agents into the same conversations your teams are already having, so what one person figures out doesn’t stay stuck with them. By making Slack the front door to your software stack, you give your workforce an open, secure work operating system where humans and agents build, execute, and succeed together.