Introduction to Agentic AI

Agentic AI is a new era of artificial intelligence that enables agents to reason, act, and collaborate, bridging the gap between knowledge and outcomes. According to the Azure blog, agentic AI is a game-changer for enterprise automation, allowing agents to interact directly with enterprise systems, retrieve data, call APIs, trigger workflows, and execute transactions. For example, Fujitsu transformed its sales proposal process using specialized agents, reducing production time by 67%. (Source: azure.microsoft.com)

Patterns of Agentic AI

There are five patterns of agentic AI that are foundational building blocks for enterprise automation. These patterns include the tool use pattern, reflection pattern, and others. The tool use pattern allows agents to interact directly with enterprise systems, while the reflection pattern enables agents to assess and improve their own outputs. For instance, GitHub Copilot relies on internal testing and refinement before sharing outputs, reducing errors and giving enterprises confidence in AI-driven processes. (Source: azure.microsoft.com)

Microsoft Foundry and Agentic AI

Microsoft Foundry is a unified, interoperable AI platform that enables developers to build, optimize, and govern AI apps and agents. Foundry provides access to over 11,000 models, including foundational, open, reasoning, multimodal, and industry-specific models. It also allows for customization and upgrading of models with minimal coding, as well as real-time model routing for high-quality results and minimized costs. Additionally, Foundry enables the creation of action-oriented, context-aware agents that automate complex business processes. (Source: azure.microsoft.com)

Summary

  • Evaluate Microsoft Foundry for building and optimizing AI apps and agents that understand business context and deliver business impact.
  • Adopt agentic AI patterns, such as the tool use pattern and reflection pattern, to enable agents to interact with enterprise systems and improve their own outputs.
  • Explore the use of LangChain, CrewAI, and LlamaIndex for seamless hosting and integration of open-source agent frameworks in Microsoft Foundry.
  • Verify the security and governance of AI systems, including customer-managed encryption, protections for data while in use, and restrictions on operator access.
  • Migrate to Microsoft Foundry for a unified, interoperable AI platform that enables fleetwide security and governance in a unified portal. (Source: azure.microsoft.com)