Anthropic and OpenAI introduced new products on Thursday that center on a shared concept: users should oversee groups of AI agents rather than interact with a solitary assistant. These releases reflect a broader industry transition from AI as a conversational partner to AI as a managed labor force. This development coincides with a week where the idea of AI agents reportedly contributed to a $285 billion decline in software stock values.
The effectiveness of this supervisory approach is still uncertain. Existing AI agents often need significant human oversight to correct mistakes, and no external assessment has verified that multi-agent systems consistently deliver better results than a single human developer working independently. Despite these challenges, both companies are heavily investing in agent technology.
Anthropic has rolled out Claude Opus 4.6, an updated iteration of its top-tier AI model, integrated with a functionality named “agent teams” within Claude Code. This feature allows developers to initiate multiple AI agents that partition a task into separate components, operate autonomously in coordination, and execute simultaneously. In practical terms, agent teams present as a divided terminal interface: developers can navigate between subagents using Shift+Up/Down keys, assume direct control over any individual agent, and observe the others continuing their work.
Anthropic characterizes this tool as most effective for “tasks that split into independent, read-heavy work like codebase reviews.” It is currently accessible as a research preview.
OpenAI has launched Frontier, an enterprise platform designed to “hire AI co-workers who take on many of the tasks people already do on a computer.” Frontier provides each AI agent with a distinct identity, access permissions, and memory, and it interfaces with established business systems such as customer relationship management software, ticketing tools, and data storage solutions.
Barret Zoph, OpenAI’s general manager of business-to-business, stated to CNBC, “What we’re fundamentally doing is basically transitioning agents into true AI co-workers.”
