Microsoft Azure AI Foundry Agent Service: Deep Research in public preview

Effective July 7, 2025, Microsoft has introduced a powerful new capability in Azure AI Foundry: Deep Research, now available in public preview (registration necessary). You can automate web-scale research tasks using programmable AI agents with this new service.

With Deep Research, you can build agents that plan, analyze, and synthesize information from across the web, generate transparent and auditable outputs, and integrate into your business workflows.

 

 

What is Azure AI Foundry Agent Service?

Azure AI Foundry Agent Service is a platform from Microsoft that helps you run AI agents. It connects models, tools, and frameworks into a unified system for operating AI agents. These agents can make decisions, invoke tools, and complete tasks independently or in coordination with other agents or humans.

Reminder: Microsoft now differentiates between “AI agents“/”agents”, meaning autonomous agents, and “Customer service representative“/”service representative”/”representative”, meaning humans: https://www.schneider.im/microsoft-entra-agent-id-a-new-era-of-identity-for-ai-agents/.

 

What is Deep Research?

While classic large language models (LLMs) focus on generating fluent responses based on statistical patterns in text, reasoning LLMs are designed to break down complex tasks, follow multi-step logic, and interact with tools.

Deep Research is a programmable agent that can be embedded into apps, workflows, or other agents to perform research automatically. It uses a reasoning LLM, called o3-deep-research.

It is built for tasks that require deep analysis, synthesis, and traceable insights from online sources.

 

Key Features

  • Automates complex research: Instead of manually searching, reading, and summarizing, Deep Research agents do it for you—step by step.
  • Produces traceable results: Every insight includes citations and a documented reasoning path, which is essential for compliance, reporting, and decision-making.
  • Integrates easily: You can trigger Deep Research from your business apps, dashboards, or workflows—making it part of your daily operations.
  • Scales with your needs: Whether you need weekly market analysis or real-time regulatory updates, Deep Research can be reused and adapted.

 

Licensing & Cost

Deep Research uses the o3-deep-research model, and pricing 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 Bing Search grounding and GPT model usage.

 

Need Help with Licensing or Integration?

If you’re considering Deep Research in your Azure operations, SCHNEIDER IT MANAGEMENT can help you evaluate licensing options, integration strategies, and deployment planning. Contact us to discuss how this new capability can support your business goals.

 

More Information

Please contact us for expert services on your specific Microsoft software and online services requirements and to request a quote today.

Questions about licensing?
Our experts have answers.

Share article

Related articles

SCHNEIDER IT MANAGEMENT 2016-10-04 Logo Standard 480 x 480
Newsroom
Find the latest licensing
news and updates.