Building a smarter research agent just got easier for businesses as Microsoft unveiled the public preview of public preview of Deep Research for Azure AI Foundry.
This latest addition taps OpenAI’s most sophisticated models and gives developers a toolset that shifts research automation into high gear for complex, enterprise environments.
Deep Research is not a digital assistant you fire up for quick answers. Instead, it is a fully programmable framework accessible by API and software kit, designed to fit within any business system or workflow.
Developers can create agents that understand intent, plan research, and collect information from the freshest, highest quality sources across the web. Every insight, every step, and every source is transparently documented, enabling compliance and deep trust.
The core engine, known as the o3 deep research model, brings context with Bing Search to ensure it only draws on current and reliable content. Each research question is refined by the AI to clarify the user’s real need before the system sources and synthesizes data. Results are then not only contextual but also traceable by design.
Programmable Agents for the Enterprise
Traditional tools like ChatGPT offer impressive chat-based intelligence, but Deep Research aims higher: embedding, streamlining, and governing automated research at scale inside enterprise systems. The process does not end with information gathering. These agents learn, pivot, and adapt their research path as new findings surface, and their final structured reports include every step and source along the way.
For programmatic minds, the flexibility is the real draw. You can set a research agent to trigger as part of a workflow chain. Imagine one agent digging deep on a market analysis, another transforming the findings into a polished presentation, and a third sending key results to decision makers — all seamlessly orchestrated.
“It is all about giving businesses control and transparency without handcuffing them to one workflow,” said a Microsoft spokesperson.
Security and governance are built-in, so organizations can meet strict compliance standards while making research part of daily digital operations. This opens doors for regulated industries that demand full audit trails or for teams where transparency and accuracy have no room for compromise.
Pricing has been set per million tokens, with clear rates for input, cached input, and output. Use of additional web search and base models follows their separate charges, ensuring enterprises see exactly what they spend.
The push for programmable, composable research comes as businesses look for new ways to automate knowledge work. With Deep Research, they finally get a research agent that does not just answer questions, but becomes an intelligent, auditable building block.
Sign ups for the public preview are open now for Azure AI Foundry users, and early adopters can tap into documentation and learning modules to build their first custom agents. Microsoft promises more features and customer stories are on the horizon as this technology expands.
