OpenAI and Anthropic have separately introduced multi-agent autonomous features aimed at enterprise customers, pushing the boundaries of how businesses can deploy AI for complex tasks. The announcements, which arrived without a joint timeline, signal that both companies see autonomous agent collaboration as the next frontier in enterprise AI.
What the features do
Multi-agent systems allow several AI agents to work together autonomously on workflows that would normally require human coordination. Instead of a single chatbot answering a question, a team of agents can break down a project, assign subtasks, and exchange information without constant human prompting. Both OpenAI and Anthropic are now offering this capability to their enterprise clients, though they have not disclosed specific pricing or availability details.
The companies did not release technical specifications or case studies alongside the unveiling, but the move aligns with a broader industry push toward agents that can operate independently over longer periods.
For businesses, autonomous multi-agent features promise to automate more complex processes — supply chain coordination, code review pipelines, or customer support escalation chains — where a single model might struggle to handle the full scope. Enterprise customers can now configure teams of AI agents with distinct roles, potentially reducing the need for manual handoffs between software tools.
The competition between OpenAI and Anthropic in this space is tight. Both firms have been racing to demonstrate that their models can handle real-world enterprise workloads reliably. By launching multi-agent features around the same time, they are effectively challenging each other's claims about which system is more suitable for autonomous business operations.
Unanswered questions about safety and control
Autonomous agents raise obvious questions about oversight. If multiple AI agents act without human approval at each step, how do companies ensure they stay within guardrails? Neither OpenAI nor Anthropic has detailed the safety mechanisms built into their new features. Observers expect both firms to publish documentation or white papers in the coming weeks as early adopters begin testing.
The rollout also reignites debate over how much autonomy to grant AI in business-critical systems. Some enterprise clients may welcome the efficiency gains; others will want strict limits on what the agents can do without a human sign-off.
The features are now available to select enterprise customers, and broader access is expected later this year. How much control those customers will keep — and how the agents handle errors — remains a question the companies have not yet answered.
