OpenAI and Anthropic unveil multi-agent autonomous features for enterprise use

The two most valuable AI companies on the planet just made the same bet at the same time. OpenAI and Anthropic have both rolled out multi-agent platforms designed to let enterprises automate complex workflows, essentially turning AI from a chat assistant into an autonomous workforce that can operate across phones, desktops, and internal systems.

The timing isn’t a coincidence. Both companies are positioning their agent platforms as the default infrastructure layer for enterprise automation, a market that could dwarf their current consumer-facing revenue streams.

What each company actually shipped

Anthropic launched Claude Managed Agents in public beta on April 8, 2026. The product lets enterprises define tasks, integrate tools, and deploy long-running agents without needing external orchestration frameworks. In plain English: companies can set up AI workers that handle multi-step processes, like onboarding a new employee or triaging engineering tickets, without stitching together a bunch of third-party tools to make it work.

By May 2026, Anthropic pushed its multi-agent orchestration capabilities into public beta alongside a novel feature called “dreaming.” The dreaming capability allows agents to learn and improve from previous interactions, essentially letting them get better at their jobs over time without explicit retraining.

OpenAI took a slightly different path. Its Frontier enterprise platform launched in February 2026, built around shared context and permissions for AI agents. Then in April 2026, OpenAI followed up with Workspace Agents, a no-code alternative to custom GPTs that lets non-technical teams deploy agents without writing a single line of code.

Anthropic reportedly drew from dozens of customer implementations to shape its product. OpenAI leaned on strategies from seven frontier companies for its enterprise design.

Where the agents are actually working

Software engineering activities account for nearly 50% of agent activity on Anthropic’s public API as of early 2026. That’s a staggering concentration in a single use case, suggesting that coding tasks, from writing boilerplate to debugging to code review, represent the lowest-hanging fruit for autonomous agents.

But both companies are pushing well beyond engineering. Their enterprise platforms target finance, human resources, and project management workflows. The vision is agents that can process expense reports, draft contract amendments, schedule cross-team standups, and flag compliance issues, all running in the background while humans focus on higher-level decisions.

Both companies are embedding comprehensive tools directly into their platforms, a strategic choice that reduces reliance on specialized competitors in the agent orchestration space.

What this means for the broader market

The broader industry trend toward multi-agent systems, where multiple AI agents collaborate on complex tasks rather than a single model handling everything, is reshaping the competitive landscape. Integrated model solutions that combine orchestration with runtime capabilities are gaining traction over third-party agent frameworks, which could squeeze margins for middleware providers.

Anthropic’s dreaming feature, if it delivers meaningful performance improvements over time, could create a compounding advantage where agents deployed earlier simply get better faster. OpenAI’s no-code Workspace Agents, meanwhile, lower the barrier to adoption dramatically, potentially winning on distribution even if the underlying capabilities are comparable.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.