At the World Artificial Intelligence Conference (WAIC) in Shanghai on July 21, 2026, Alibaba Cloud announced a significant shift in how it positions its entire cloud infrastructure. The company is moving from an “AI-native” architecture to an “agent-native” one — and it launched three products to back that up.

The suite, called Agent Native Cloud, includes AgentLoop, AgentTeams, and the existing AgentRun platform, now expanded. Together they form a lifecycle management stack for enterprise AI agents: build, run, monitor, optimize, and coordinate.

What Alibaba Actually Launched

AgentRun covers the core lifecycle — development, deployment, and operations of AI agents. Think of it as the runtime layer.

AgentLoop adds observability: real-time tracing, evaluation, and performance optimization for individual agents as they work. This is the piece most enterprises are missing right now. Running an agent is straightforward. Knowing why it failed, how to fix it, and how well it performed across thousands of runs is the hard part.

AgentTeams goes a level above: coordination and governance for fleets of agents working on complex workflows. Multiple specialized agents, managed together, with human oversight baked in.

The company also introduced Agentic Computer, a secure cloud-based execution environment, and TokenWorks, a new inference service via its Platform for AI.

The Numbers That Matter

Alibaba didn’t just announce products — it showed internal results. Fifteen coordinated AI agents now handle 85% of the company’s own developer support requests. Operational support time has dropped by 90%. Software release cycles have been compressed to a single day.

Those aren’t pilot numbers. That’s a production deployment at one of the world’s largest technology companies.

The shift from managing one AI model to managing a coordinated team of AI agents is where most businesses are struggling right now. The tools have existed for individual agents. The infrastructure for running agent teams reliably, at scale, with audit trails and governance, has not.

That’s the gap Agent Native Cloud is designed to close.

What This Means for Business

The direction of enterprise AI is becoming clearer. The early wave was about adding AI features to existing software. The current wave is about replacing entire workflows with coordinated agent fleets. Alibaba is architecting its cloud around that assumption.

For businesses evaluating AI infrastructure, a few things stand out:

Governance is the feature. AgentTeams includes built-in controls for coordinating multiple agents and keeping humans in the loop. As regulatory pressure increases globally, this isn’t optional for enterprise deployments.

Observability is becoming table stakes. The reason most AI agent projects fail to scale is that businesses can’t see inside them. AgentLoop’s real-time tracing solves a problem that’s blocking a lot of enterprise rollouts right now.

The infrastructure layer is maturing fast. A year ago, building a multi-agent system required significant custom engineering. Platforms like Agent Native Cloud are industrializing that work — which means the cost of entry for serious AI automation is dropping.

The Competitive Picture

Alibaba is not alone in this space. Microsoft’s Azure AI, AWS Bedrock, and Google’s Gemini Enterprise portfolio are all building similar infrastructure for agent orchestration. But Alibaba’s WAIC announcement is notable for a few reasons: it’s backed by internal production metrics, it includes native governance tools that many competitors are still bolting on, and the “agent-native” framing signals a more fundamental architectural commitment than typical product launches.

For businesses that operate in or are entering markets where Alibaba Cloud is the dominant infrastructure provider — Southeast Asia, China, the Middle East — this is directly relevant to how you’ll build and govern AI agents over the next two years.

The Bottom Line

Managing a single AI assistant is a solved problem. Managing a workforce of AI agents — with different roles, handoffs, error recovery, and audit requirements — is not. Agent Native Cloud is Alibaba’s answer to that problem, and the production metrics suggest it works.

The question for every business leader is the same one it’s been for the past 18 months: are you building the internal capability to operate AI agent teams, or are you waiting until the tools are perfect? The tools are getting very good, very fast.

If you want to understand what running an AI agent workforce looks like in practice — and what it takes to build one that actually delivers measurable results — talk to us about Omni Ops. We’ve deployed coordinated agent systems for businesses that needed real outcomes, not demos.