Building Practical AI Features with Azure OpenAI and Azure AI Search

AI demos are easy. Building AI features that fit into existing applications; and choosing the right platform to build them on; is much harder. In this full-day, hands-on workshop, participants build a small but realistic support-agent workflow using Azure OpenAI and Azure AI Search. Starting from a prepared codebase, attendees implement core capabilities themselves: calling large language models from backend services, enforcing structured outputs, and combining model responses with enterprise search results. Beyond the hands-on build, the workshop explicitly addresses a common source of confusion in the Microsoft ecosystem: there are multiple valid ways to build agent-like solutions. Azure AI infrastructure, Microsoft AI Foundry, and Copilot Studio all solve different problems; and using the wrong one often leads to stalled projects or unnecessary complexity. Throughout the day, we map concrete implementation choices to platform options and discuss when full control is required, when managed services make sense, and when a SaaS-style approach is the right trade-off. Participants leave with reusable architectural patterns and practical guidance for choosing the right Microsoft AI platform for real-world solutions.

Agenda:

• Problem framing and architecture overview
• The Microsoft AI landscape; Azure, Microsoft AI Foundry, and Copilot Studio
  • Control vs speed
  • Custom code vs managed services
  • When each option breaks down
• Deploying the baseline Azure infrastructure
• Calling Azure OpenAI from backend services
• Enforcing structured outputs and validation
• Integrating Azure AI Search for retrieval-augmented responses
• Observability, error handling, and predictable behavior
  • Optional extension: Exposing the agent through Copilot Studio
• Wrap-up; decision guidelines, common pitfalls, and next steps

Target audience:

Software architects and developers with Azure experience who design, build, or review AI-enabled applications.