Microsoft has introduced two new additions to Azure AI Foundry, a cohesive environment that combines Microsoft’s AI tools with OpenAI’s foundation models to support design and management of AI apps and agents.
These important innovations empower “businesses to reimagine AI not just as an assistant — but as an active digital workforce,” according to Steve Sweetman, Azure OpenAI Service Product Lead.
Responses API and Computer-Using Agent (CUA) will help users scale their agentic AI initiatives more quickly and efficiently, with potential applications in a wide range of functions including customer service, IT operations, finance, and supply chain management.
Sweetman describes Responses API as “the new foundation for leveraging Azure OpenAI Service’s powerful built-in tools, combining the simplicity of the Chat Completions API with the advanced capabilities available through Assistants API and Azure AI Agent Service.”
In less brand-focused terms, developers can now instruct AI to interact with multiple tools through a single API call without losing context.
Simplified Agentic AI Development
Responses API includes a computer-use tool for automation and software interactions, a file search feature for data interactions and retrieval, and function calling capabilities. Interactions are linked using response IDs, and the whole process is secured through Azure’s enterprise-grade data privacy architecture. Microsoft says that by consolidating these features, it is simplifying agentic AI development.
A specialized AI model in Azure OpenAI Service, CUA enables AI to interact with GUIs, navigate applications, and automate actions using natural language prompts. CUA’s automation does not require scripted instructions. Instead, the technology can adapt and perform actions based on the visual information displayed on the screen.
AI Agent & Copilot Summit is an AI-first event to define opportunities, impact, and outcomes with Microsoft Copilot and agents. Building on its 2025 success, the 2026 event takes place March 17-19 in San Diego. Get more details.
CUA can autonomously interact with a user interface by opening and closing applications, selecting options, and filling out forms while adapting to changes in the UI. It can operate across web and desktop applications and integrate multiple systems without needing an API. All these capabilities are actioned through natural language prompts.
The Responses API simplifies AI agent development by coordinating various AI tools in an automated pipeline. Its efficiency and scalability make it ideal for enterprise use cases.
However, for me, CUA represents one of the most significant advancements in agentic AI capabilities, moving beyond the development opportunities provided by Responses API. Just recently, we reported on Microsoft’s breakthrough foundational model for multimodal agents, Magma.
Although the implementation of Magma-like features has accelerated and the interactive capabilities of a digital workforce are being made available, there is still a human in the loop. And, so far, we are only discussing operations within digital software environments.
This human-in-the-loop concept is crucial as it ensures AI is aligned with business goals and values. It also provides a layer of oversight and control, particularly in sensitive operations. Sweetman confirmed this point: “Automated and human-in-the-loop detection systems monitor execution patterns, identifying anomalous behaviors and enforcing governance policies.”
It is encouraging to see a focus on security and trustworthy AI in the development of the Azure AI Foundry. Moving forward, this commitment to safety and reliability should also encompass job security as this technology rapidly develops, making the wider public feel confident about the direction of travel.
