Every Monday your team talks about automating things. "We should set up a workflow for that." Everyone nods. The standup ends. Nobody opens the automation tool. The conversation about automation remains the most manual ritual in your company.
The problem was never capability. Zapier has 8,000 integrations. Make has visual flows. n8n lets you self-host. The tools exist — but they live in a separate tab, behind a separate login, in a separate mental context. And context-switching is where good intentions go to quietly decompose.
Two announcements in the past two weeks confirmed a pattern that's been building all month.
On March 19, Google rolled out Workspace Studio to all business customers. It's a no-code platform where employees describe what they want automated in plain language, and Gemini 3 — Google's latest AI model — builds the agent. No separate tool. No new login. You type "every Friday, ping me to update my tracker" directly inside the workspace you already have open. During the alpha, customers completed over 20 million tasks through these agents. Kärcher, the cleaning equipment manufacturer, chained multiple agents together and cut feature-drafting time by 90% — from hours of manual consolidation to a ready-to-review plan in two minutes.
Five days later, on March 24, Oracle launched 22 Fusion Agentic Applications — production-ready AI agents built natively into its cloud suite for supply chain, procurement, and finance. Not a separate AI add-on. Native.
Google and Oracle didn't arrive at this independently. Earlier in March, Microsoft and Slack laid the same groundwork. On March 9, Microsoft announced Copilot Cowork — an autonomous agent (a bot that takes actions on its own, not just answers questions) built on Anthropic's Claude, executing multi-step workflows across Outlook, Teams, Excel, and PowerPoint. On March 13, Slack published its "Agentic Productivity" vision, positioning chat as the hub where Salesforce's Agentforce bots handle CRM lookups, IT tickets, and project summaries — triggered by a message in the channel where you were already complaining about doing it manually. Four platforms, one month, identical strategy: embed the automation engine where users already live.
The infrastructure backs the pattern. Slack shipped a native MCP server — MCP (Model Context Protocol) is a universal plug standard that lets AI tools connect to your apps, like USB but for data. Since October, Slack reported a 25x increase in both search queries and tool calls from third-party agents. Over 50 partners, including Anthropic, Google, and OpenAI, now build agents that live inside Slack. MCP hit 97 million monthly SDK downloads in March, up from 2 million at launch in November 2024. No longer experimental.
Now the price tag. Microsoft's new E7 tier costs $99 per user per month (general availability on May 1, 2026). Agentforce in Slack requires a separate Salesforce license — the agents are free in Slack, but the brain behind them isn't. Google Workspace Studio ships with existing business plans, making it the most accessible entry point, but Google's history of sunsetting products makes long-term bets uncomfortable. And letting AI execute workflows from a natural-language prompt is exactly one sloppy description away from a production incident. No independent study has confirmed whether embedded automation drives higher completion rates than standalone tools.
The real lesson is quieter than the announcements. For years, we treated automation adoption as a capability problem — "the tool needs more integrations, more features, more power." It was a distribution problem. Put the automation where people already spend their day, and they actually use it. Same reason the best note-taking app is the one already open on your screen ⚙️
Standalone automation dashboards aren't dead — they're becoming the backend that agents call. But if your automation requires someone to leave the conversation to set it up, you've already lost them 🧘
