The tech world took big steps this week toward AI agents that work together across platforms, breaking down old barriers between different systems. Microsoft made headlines by adopting Google’s Agent2Agent (A2A) protocol, an open standard for letting AI agents share goals and actions securely. By adding A2A to Azure AI Foundry and Copilot Studio, Microsoft ensures its agents can now collaborate with external ones—even those built on Google’s tools—to handle complex workflows like scheduling meetings or managing emails. Microsoft also joined Google’s A2A working group on GitHub, signaling a rare collaboration between rivals to shape the future of interoperable AI.

IBM unveiled its vision for enterprise-ready AI agents at its Think 2025 conference, launching watsonx Orchestrate, a platform that claims to build agents in under five minutes. With 150+ prebuilt agents and integrations for apps like Salesforce and SAP, IBM aims to tackle the “fragmentation” that slows down AI projects in big companies. They also highlighted TufinMate, a security-focused agent that lets network engineers manage configurations via Slack or Teams—showing how agents can simplify highly technical tasks. According to IBM, orchestration tools are the “secret weapon” to unlock the full economic potential of AI, estimated at $4.4 trillion annually.

AWS and IBM deepened their partnership to advance agentic AI, combining IBM’s automation tools with AWS’s cloud infrastructure. Their joint effort focuses on creating agents that can reason independently, like automatically resolving IT outages by analyzing logs across hybrid clouds. This aligns with industry trends where multi-cloud agent systems are becoming critical for businesses juggling apps across Microsoft Azure, Google Cloud, and AWS.

In the startup world, San Francisco’s AI Agent Founders Gatherings showcased tools for low-code agent orchestration, while Microsoft’s Agents 101 Hackathon taught developers to build adaptive agents using A2A. These events highlighted a growing emphasis on making agent collaboration accessible, even to non-experts.

As companies rush to adopt agentic AI, challenges remain—like ensuring security when agents share data across platforms. However, this week’s moves by Microsoft, IBM, and Google suggest a future where AI teams up like human coworkers, regardless of which company built them.