Introducing Deep Research in Azure AI Foundry Agent Service
We’re thrilled to introduce the public preview of Deep Research in Azure AI Foundry—an API and SDK-powered solution showcasing OpenAI’s innovative research capabilities.
Transform Your Web Research Automation
Today marks an exciting milestone with the public preview of Deep Research in Azure AI Foundry. This powerful service combines an API and software development kit (SDK) to leverage OpenAI’s sophisticated research technology, seamlessly integrated into Azure’s robust platform for enterprises.
With Deep Research, developers can create intelligent agents that meticulously plan, analyse, and synthesise web data. This enables the automation of complex research tasks, the production of transparent and verifiable outputs, and the effortless creation of multi-step workflows alongside other Azure AI Foundry tools and agents.
AI Agents and Knowledge Work: The Future of Research Automation
Thanks to generative AI and large language models, research and analysis has become quicker and more efficient than ever. Tools like ChatGPT Deep Research and the Researcher in Microsoft 365 Copilot are enhancing productivity and document workflows for countless users.
As organisations aim to take research to the next level—by embedding deep research into their business applications, automating intricate processes, and managing information on an enterprise scale—the need for programmable and verifiable research automation becomes evident.
Enter Azure AI Foundry and Deep Research: they provide the flexibility to implement, extend, and orchestrate top-notch research as a service throughout your entire enterprise, connecting it directly to your data and systems.
Unleashing Deep Research Potential in Azure AI Foundry
Deep Research in Foundry Agent Service is tailored for developers who want to go beyond traditional chat interfaces. By offering Deep Research as a customizable agent tool via API and SDK, Azure AI Foundry allows customers to:
- Automate Large-Scale Research: Utilise a top-notch research model powered by Bing Search, ensuring that every insight is traceable and supported by credible sources.
- Create Programmable Agents: These agents can be triggered by applications, workflows, or other agents, transforming deep research into a reusable, production-ready service.
- Manage Complex Workflows: Combine Deep Research agents with Logic Apps, Azure Functions, and additional connectors from Foundry Agent Service to automate tasks such as reporting and notifications.
- Ensure Governance at an Enterprise Level: With Azure AI Foundry’s focus on security, compliance, and observability, customers gain full control and visibility over how research is conducted and utilized.
Unlike conventional chat assistants, Deep Research in Foundry Agent Service is designed to evolve with your requirements, ready for automation, expansion, and integration with future data sources.
How It Works: Architecture and Agent Workflow
Deep Research in Foundry Agent Service is built for flexibility, transparency, and composability, ensuring your research automation meets robust business demands.
At its core, the Deep Research model, o3-deep-research, orchestrates a multi-step research pipeline that is tightly aligned with Grounding with Bing Search and employs the latest OpenAI models:
- Understanding Intent and Defining the Task:
When a user or an application presents a research query, the agent uses GPT-series models, including GPT-4o and GPT-4.1, to refine the question, collect extra context if required, and precisely define the research scope. This ensures the output is relevant and actionable, optimising searches based on your specific business situation. - Web Grounding Using Bing Search:
Upon task definition, the agent securely engages the Grounding with Bing Search tool to gather a curated set of reliable, up-to-date web data. This guarantees that the research model operates from authoritative, current sources, eliminating outdated or irrelevant content. - Executing Deep Research Tasks:
The o3-deep-research model initiates the research task execution, engaging in critical thinking, analysis, and synthesis of information across all identified sources. Unlike basic summarising, it reasons step-by-step, adapts to new insights, and crafts a comprehensive answer that considers nuances, ambiguities, and emerging patterns in the data. - Transparency, Safety, and Compliance:
The final result is a structured report that not only provides the answer but also outlines the model’s reasoning process, citation sources, and any clarifications made during the session. This level of detail ensures that every answer is fully auditable—a crucial feature for regulated industries and high-stakes situations. - Seamless Integration:
With Deep Research accessible through an API, Azure AI Foundry allows you to invoke research from any platform—be it custom business applications, internal portals, workflow automation tools, or as part of a broader agent ecosystem. For instance, one agent could perform in-depth web analysis, while another generates a slide presentation with Azure Functions, and a third dispatches results to decision-makers via Azure Logic Apps. This composability transforms research from a one-off task into a foundational building block for digital transformation and ongoing intelligence.
The adaptable architecture of Deep Research means it can be smoothly incorporated into diverse enterprise workflows and applications. Already, businesses across different sectors are exploring how these programmable research agents can enhance high-value tasks—from market analysis and competitive intelligence to extensive analytics and regulatory reporting.
The pricing for Deep Research (model: o3-deep-research) is structured as follows:
- Input: $10.00 per 1 million tokens.
- Cached Input: $2.50 per 1 million tokens.
- Output: $40.00 per 1 million tokens.
Be aware that search context tokens will incur input token charges based on the model in use. Additionally, there will be separate fees for Grounding with Bing Search and the underlying GPT model used for clarifying questions.
Start Your Journey with Deep Research
Deep Research is now accessible in a limited public preview for Azure AI Foundry Agent Service clients. Here’s how to get started:
We’re eager to see the innovative solutions you’ll develop. Keep an eye out for customer stories, new features, and future improvements that will further unveil the next generation of enterprise AI agents.
