Microsoft are excited to announce the public preview of Deep Research in Azure AI Foundry—an API and software development kit (SDK)-based offering of OpenAI’s advanced agentic research capability, entirely integrated with Azure’s enterprise-grade platform. This innovative tool allows developers to build agents capable of deeply planning, analyzing, and synthesizing information from across the web. It enables the automation of intricate research tasks, ensuring outputs are transparent and auditable.
AI Agents and Knowledge Work
The advent of generative AI and large language models has revolutionized the pace of research and analysis, with solutions like ChatGPT Deep Research and Researcher in Microsoft 365 Copilot transitioning productivity to new heights. Organizations are now focusing on embedding deep research capabilities into their business applications, automating multifaceted processes, and ensuring governance at enterprise scale. This transformative shift underscores the imperative need for research automation that is programmable, composable, and auditable.
Deep Research Capabilities in Azure AI Foundry Agent Service
Deep Research in Foundry Agent Service is designed for developers ready to advance beyond basic chat interfaces. By leveraging a composable agent tool via API and SDK, Azure AI Foundry empowers clients to:
- Automate web-scale research using a premier research model integrated with Bing Search, ensuring every insight is traceable and backed by credible sources.
- Programmatically develop agents usable in apps, workflows, or other agents, effectively converting deep research into a reusable and production-level service.
- Seamlessly orchestrate intricate workflows: Combine Deep Research agents with Logic Apps and Azure Functions to automate reporting, notifications, and beyond.
- Guarantee enterprise governance: Azure AI Foundry provides comprehensive security, compliance, and observability, ensuring full control over the research processes.
Contrasting standard chat assistants, Deep Research in Foundry Agent Service is adaptable to evolving client needs, prepared for automation, extensibility, and integration with future internal data sources.
How It Works: Architecture and Agent Flow
The architecture of Deep Research in Foundry Agent Service emphasizes flexibility, transparency, and composability to enable robust research automation consistent with business demands.
The core Deep Research model, o3-deep-research, orchestrates a multi-step research pipeline that is closely integrated with Grounding using Bing Search and utilizes advanced OpenAI models. Here’s a breakdown of the process:
- Clarifying Intent and Scoping the Task: Upon a research query submission, the agent employs GPT-series models to clarify the question and establish precise task scope, ensuring optimized search results for business scenarios.
- Web Grounding with Bing Search: The agent invokes the Grounding with Bing Search tool to collect a curated set of high-quality web data, laying a solid foundation of reliable, current sources.
- Deep Research Task Execution: The o3-deep-research model processes the research task, analyzing and synthesizing information methodically across all identified sources. This nuanced method yields comprehensive answers while adapting to newly discovered insights.
- Transparency, Safety, and Compliance: The output is a structured report detailing not just answers but the rationale and source citations used throughout the research process, providing audit trails crucial for industries requiring strict compliance.
- Programmatic Integration and Composition: By allowing Deep Research to be exposed as an API, Azure AI Foundry enables integration of research into custom business apps and internal portals, facilitating automation of multi-agent workflows.
This adaptable architecture enables Deep Research to be embedded seamlessly across varied enterprise applications, facilitating significant operational efficiency.
Pricing for Deep Research
The pricing structure for Deep Research (model: o3-deep-research) is as follows:
- Input: $10.00 per 1M tokens.
- Cached Input: $2.50 per 1M tokens.
- Output: $40.00 per 1M tokens.
Additional charges apply for Grounding with Bing Search and the base GPT model used for clarification queries.
Get Started with Deep Research
Deep Research is currently accessible in limited public preview for Azure AI Foundry Agent Service customers. To embark on your research automation journey:
- Sign up for the limited public preview to gain early access.
- Explore our documentation for detailed insight about the feature.
- Engage with learning modules and build your first agent with Azure AI Foundry.
- Begin developing your agents today in Azure AI Foundry.
