On a single day, the world’s most powerful technology firms remade the purpose of artificial intelligence. The era of the AI ‘copilot’—a tool that assists—gave way to the age of the AI ‘agent,’ a tool that acts. Microsoft reconfigured its software to run on rival models, while OpenAI and Stripe launched a protocol for agents to buy and sell goods directly. A new digital landscape was revealed, and the race is now on, not through collusion but through fierce competition, to write the rules and own the core infrastructure of this new economy.

This is Modra. A wine town in the hills above Bratislava. The night air is cool, smelling of damp earth and last month’s harvest. Here, the twenty-first century feels distant. It is not.

A Mandate to Act

The future of software changed today. It did not happen in a Silicon Valley garage. It happened on screens, in a series of quiet announcements that landed like seismic shocks. The change has a name: the agentic pivot. For years, artificial intelligence was a copilot. It sat beside you. It helped you write an email or summarize a report. Today, the industry began handing the AI the controls. The new AI is an agent. It does not just advise. It acts.

Microsoft made the first move. Its new Office Agent will not just help you analyze a spreadsheet. It will run the analysis, generate the report, and schedule the follow-up meeting. In a significant maneuver, Microsoft revealed the agent is powered not by its long-term partner OpenAI, but by rival lab Anthropic. The signal was clear: Microsoft intends to be a broker of AI services, not a vassal to one model. It will own the workflow, the new battlefield.

Then came the money. Stripe, the payments giant, and OpenAI unveiled the Agentic Commerce Protocol. They shipped it today inside ChatGPT for all U.S. users. It lets the AI buy things. A user can ask for a flight to London, and the agent can find it, book it, and pay for it, instantly. Will Gaybrick, a Stripe executive, called it “the economic infrastructure for AI.” Google has a competing protocol, but it remains a specification on paper. OpenAI and Stripe just built the first railroad.

Not everyone is racing toward autonomous action. Asana, a project management company, offered a different vision. It announced “AI Teammates,” designed for collaboration, not replacement. The philosophy is one of oversight, of a human always in the loop. This created the central tension of the day: a contest between AI as an autonomous actor and AI as an expert collaborator.

The New Battlefield

This is not just a semantic debate. It is a race to build and control a new digital landscape called the “agentic layer.” This is the territory where agents will live, transact, and operate. Whoever writes the rules for this layer, whoever controls its standards for payments and identity, stands to gain immense power. The risk for businesses and individuals is not a conspiracy. It is lock-in, a quiet dependence on one company’s way of doing things.

The pivot is happening because the old way was failing. An MIT study found 95 percent of corporate AI projects showed no return on investment. The copilots were not delivering. But a KPMG survey showed enterprise adoption of agents leaping from 11 percent to 42 percent in a single quarter. Early results are driving the shift. One startup, Maximor, builds finance agents. It claims a real estate client, Rently, cut its deal-closing process from eight days to four. The agents did the work.

The Best Obtainable Truth

The facts, laid out in internal industry analysis, show a clear and decisive turn. There is no evidence of collusion. There is only evidence of a race. The questions are sharp. Who will own the rails of this new agent-driven economy? Will the protocols for commerce be open, like the web, or closed, like an app store?

Today, the technology industry stopped talking about just generating content. It started building AI that acts in the world. The blueprints were laid. The race for the agentic layer had begun.