An economist's Twitter thread is circulating in AI circles this week. Otto Zastrow, who writes on economic theory and futures, sketched out what competition between firms might look like once AI agents are calling the shots — not just handling tasks, but running strategy, allocating capital, and responding to market signals faster than any human executive could.
Zastrow's core argument is about where competitive advantage moves. Once autonomous agents are operating companies, the traditional inputs — talented people, strong management, proprietary know-how — start to matter less. What matters instead is the quality of your agentic infrastructure: compute, energy, data access, regulatory positioning. These become the new moats. And because agent strategies can compound and improve recursively, early leads could become very hard to close.
The thread also raises something stranger: what is a firm, once agents are making the decisions? If an AI can negotiate, contract, and execute with minimal human involvement, the firm stops being a cluster of human relationships and starts looking more like a node in a larger machine. Zastrow suggests the economy might end up resembling a swarm — autonomous agents cooperating and competing across what are today hard organisational boundaries.
It's speculative, but not abstract. Multi-agent orchestration platforms, AI-native startups running on skeleton crews, and enterprise automation tools are already building toward this world. For anyone tracking agentic AI, Zastrow's thread is a useful frame — and a reminder that the economic implications stretch well beyond productivity gains on existing workflows.