> Markdown version of [/videos/1532-agentic-ai-from-theory-to-practice-developing-multi-agent-ai-systems-on-azure?t=737](https://www.wearedevelopers.com/videos/1532-agentic-ai-from-theory-to-practice-developing-multi-agent-ai-systems-on-azure?t=737). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Agentic AI - From Theory to Practice: Developing Multi-Agent AI Systems on Azure

Move beyond basic chat interfaces. Learn how to build, evaluate, and scale specialized multi-agent AI workflows on Azure using Semantic Kernel and robust CI/CD evaluators.

- **Speakers:** [Ricardo](https://www.wearedevelopers.com/@ricardo)
- **Event:** World Congress 2025
- **Published:** August 20, 2025
- **Duration:** 26:35
- **URL:** https://www.wearedevelopers.com/videos/1532-agentic-ai-from-theory-to-practice-developing-multi-agent-ai-systems-on-azure

## Summary

Moving beyond basic chat interfaces, modern development focuses on multi-agent AI systems integrated directly into enterprise business processes. By pairing large language models with specific instructions and internal API tools, developers can build specialized agents designed to automate complex workflows like document generation, travel booking, or contract analysis. Microsoft's ecosystem supports this transition through a structured lifecycle encompassing rapid ideation, secure implementation, and observable long-term operations.

Azure AI Foundry serves as the core orchestration layer, providing a unified umbrella for model selection, agent services, observability, and content safety. During the initial ideation phase, developers can minimize overhead by experimenting with GitHub Models to benchmark prompts and compare LLM cost-efficiency without provisioning local resources. As development moves to implementation, evaluating prompt quality becomes imperative; replacing intuitive assumptions with concrete KPIs via programmatic evaluators in CI/CD pipelines ensures robust performance. Continuous monitoring in the operational phase relies on sampling production requests to trace tool calls, token usage, and execution latency, allowing teams to catch degradation before it impacts end-users.

Scaling to a multi-agent architecture introduces advanced orchestration patterns, such as sequential execution, concurrent tasks, system hand-offs, and dynamic group chats designed around deterministic or probabilistic workflows. Using Semantic Kernel, developers can assign distinct roles and tools to individual agents, effectively bridging pure AI operations with classic, predictable logic. Ultimately, deploying these sophisticated systems through container apps or on-premise infrastructure grants developers full control, turning previously black-box AI generation into transparent, highly scalable enterprise solutions.

**Keywords:** azure AI foundry, github models, semantic kernel, multi-agent orchestration, AI model evaluators, CI/CD pipeline integration, enterprise workflow automation, AI observability, agentic design patterns, probabilistic AI workflows, deterministic AI processes, production request sampling, model cost benchmarking, API tool integration, content safety protocol

## Chapters

1. **Shifting from chat interfaces to integrated AI agents** (00:05) — Embedding AI natively into business processes removes the need for manual chat interfaces.
1. **Core components of an AI agent architecture** (01:16) — Combining a large language model with specific instructions and tool access enables automated reasoning and execution.
1. **Resolving developer challenges in AI agent implementation** (03:06) — Identifying the right model, ensuring content safety, and maintaining observability are critical hurdles when bringing AI into production.
1. **Managing AI development with Azure AI Foundry** (04:16) — A centralized suite provides model catalogs, agent services, built-in integrations, and unified observability for enterprise AI development.
1. **Orchestrating enterprise agents via the Agent Service** (05:29) — An orchestration layer simplifies integration with existing SDKs, enterprise networks, knowledge bases, and industry protocols like MCP.
1. **Aligning AI projects across standard development phases** (06:55) — Creating successful AI products requires moving deliberately from ideation and rapid testing into robust implementation and operational monitoring.
1. **Testing AI models quickly using GitHub Models** (08:29) — A vast catalog of models can be explored and tested directly in the browser without requiring immediate infrastructure provisioning.
1. **Balancing prompt performance and model cost efficiency** (09:41) — Running systematic evaluations across different prompts, models, and datasets is essential to find the right balance between output quality and expense.
1. **Connecting agents to business tools and APIs** (12:17) — Configuring agents within Azure AI Foundry involves attaching them to internal REST APIs, existing files, and active workflow triggers.
1. **Automating quality testing through CI/CD evaluation pipelines** (13:31) — Embedding automated evaluations in deployment pipelines prevents prompt modifications or model swaps from degrading system reliability.
1. **Analyzing complex contracts using multi-agent workflows** (15:22) — Specialized agents sequentially extract, compare, and validate document compliance to accelerate manual contract reviews.
1. **Measuring AI agent performance within GitHub Workflows** (17:40) — Running nightly builds against test datasets provides concrete metrics on token usage, latency, and response quality.
1. **Sampling production traffic for end-to-end tracing observability** (19:03) — Applying performance sampling to a production subset allows teams to trace execution steps without incurring excessive overhead.
1. **Selecting communication patterns for multi-agent systems** (22:07) — Choosing between sequential, concurrent, and conversational patterns depends on whether process requirements are deterministic or probabilistic.
1. **Building scalable agent workflows using Semantic Kernel** (24:36) — Combining agent and process frameworks orchestrates complex AI logic for deployment across flexible cloud or on-premises infrastructure.

## Related Moments

- [Introduction to building real-world AI agent solutions](https://www.wearedevelopers.com/videos/1538-composable-intelligence-how-henkel-and-microsoft-are-shaping-the-agent-ecosystem) (from "Composable Intelligence: How Henkel and Microsoft Are Shaping the Agent Ecosystem")
- [Scaling operations using Azure AI Foundry tools](https://www.wearedevelopers.com/videos/1535-from-traction-to-production-maturing-your-genaiops-step-by-step) (from "From Traction to Production: Maturing your GenAIOps step by step")
- [Shifting focus from isolated models to enterprise AI systems](https://www.wearedevelopers.com/videos/100130-ai-in-production-applied-ai-enterprise-use-cases) (from "AI in Production: applied AI & enterprise use cases")
- [Introduction to distributed multi-agent systems](https://www.wearedevelopers.com/videos/1976-designing-and-deploying-distributed-multimodal-multi-agent-systems-with-google-s-ai-stac) (from "Designing and Deploying Distributed Multimodal Multi-Agent Systems with Google's AI Stac")
- [Architectural patterns for developing robust generative AI applications](https://www.wearedevelopers.com/videos/1512-building-ai-applications-with-langchain-and-node-js) (from "Building AI Applications with LangChain and Node.js")
- [Rethinking team structures around AI agent capabilities](https://www.wearedevelopers.com/videos/1539-agentic-devops-how-ai-powered-automation-transforms-software-delivery-on-github-and-azure) (from "Agentic DevOps: How AI-Powered Automation Transforms Software Delivery on GitHub and Azure")

## Related Articles

- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)
- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)
- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Principal Product Manager, Agent Platform](https://www.wearedevelopers.com/jobs/ext/277541-principal-product-manager-agent-platform) at **GitHub**
- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1442858-principal-field-architect-ai-agents) at **Twilio**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**