If the past year of AI news has been dominated by what models can say, Google used its Cloud Next 2026 conference to argue that the next chapter is about what they can do. The company consolidated its sprawling AI products into a single offering, renaming Vertex AI to the Gemini Enterprise Agent Platform and folding its Agentspace product into a unified Gemini Enterprise. Alongside it came a no code agent builder for Google Workspace, a developer platform offering more than 200 models including third party options such as Anthropic's Claude, a web browsing agent called Project Mariner, and a production grade Agent2Agent protocol designed to let agents from different vendors talk to one another.
When the customer counts agents in the hundreds
The most telling detail was not a feature but a number. GE Appliances said it is deploying more than 800 AI agents built on Gemini Enterprise across its manufacturing, logistics and supply chain operations. That is a scale of automation that no longer fits the familiar story of a single chatbot bolted onto a website. It describes a workforce of software processes running alongside human staff, each handling a slice of work that used to require a person to start, check or finish.
Value, not just demonstrations
Other customers put figures on the return. Highmark Health said its Sidekick assistant delivered 27.9 million dollars in value in 2025 alone by automating research protocols and acting as a search proxy for internal teams. Deutsche Telekom described a multi agent system called MINDR, built on Gemini models, that it says cuts event management times by more than 95 percent. These are the kinds of efficiency claims that, repeated across an industry, translate directly into questions about how many people a process actually needs.
The shift inside the org chart
Google framed the moment with a forecast that agentic capabilities will move from less than 1 percent of business applications in 2024 to being embedded in roughly a third of them by 2028. If that proves even partly right, the change to white collar work will not arrive as a single dramatic layoff but as a steady redrawing of who does which task. The optimistic reading is that staff shift from tedious work to higher impact work. The harder question, unanswered at any vendor keynote, is how many of today's roles survive that shift intact.
Why this keynote matters for workers
Conference announcements are easy to discount as marketing, but the agentic pivot is now shared across Google, Microsoft, OpenAI and Anthropic, and it is backed by named customers reporting real deployments. For workers, the signal is that the AI jobs debate is moving past speculation about capability and into the practical business of rolling agents out at scale. The roles most exposed are the multi step administrative and operational tasks that these agents are explicitly built to absorb.
