Skip to main content
Ace Skills Consulting
Microsoft logoIntermediate

AI-103: Azure AI Apps and Agents Developer

Build, evaluate and ship generative AI applications and agents on Azure, and prepare for the AI-103 exam.

  • 32 hours of live, code-along sessions
  • Builds a grounded chat app and a tool-using agent
  • Evaluation, safety and monitoring treated as core skills
  • Vision, language and document intelligence in AI solutions
  • Successor path for learners who planned to take AI-102
Aligned to Microsoft exam AI-103
Duration
4 Days | 32 Hrs
Next batch
Batches on request
Delivery
Live Online · Onsite · Corporate
Level
Intermediate
Exam
AI-103
Language
English

Why learn AI-103 with Ace Skills

  • Agents built properly

    You design agents with clear instructions, scoped tools and human checkpoints, and learn why each guardrail matters.

  • Quality you can measure

    Evaluation datasets and metrics turn “the answers look fine” into evidence you can share with stakeholders.

  • Grounded in your data

    Chunking, embeddings and hybrid search are tuned hands-on so answers cite the right source documents.

  • Ready for the AI-102 successor

    If you were preparing for AI-102, the course carries over its AI services knowledge and adds the newer agent-focused skills.

Course overview

Building a prototype chatbot takes an afternoon. Building an AI application that gives grounded answers, uses tools safely, can be evaluated objectively and is monitored in production takes engineering discipline. This four-day course teaches that discipline using Azure AI Foundry and the SDKs developers use day to day.

Day one sets up a Foundry project, compares models, deploys them and works with prompts, structured output and function calling from code. Day two builds retrieval-augmented generation over your own documents with Azure AI Search, covering chunking, embeddings, hybrid search and citations.

Day three focuses on agents: designing an agent with instructions, tools and knowledge, connecting it to APIs and data, orchestrating multiple agents, and keeping a human in the loop where decisions matter. Day four rounds out the solution with vision, language and document intelligence capabilities, then covers evaluation, content safety, tracing and cost monitoring before a session on exam technique.

Labs are written in Python, with C# alternatives for the core exercises. Each lab extends the same solution, so you finish with a grounded assistant and an agent that you understand line by line.

This course is aligned to the AI-103 exam, reported as the successor to AI-102, which retired on 30 June 2026. As the exam is new, check its current outline on Microsoft Learn before booking; we update the course as details are confirmed.

Key features

  • Code-along labs in Python with C# alternatives
  • End-to-end RAG solution with Azure AI Search
  • Agents with tools, knowledge and multi-agent orchestration
  • Evaluation runs and content safety configuration
  • Tracing and monitoring of AI applications
  • Vision, language and document intelligence integration
  • 30 days of post-training doubt support

Who should attend

  • Developers building generative AI features or assistants
  • Learners who were preparing for the retired AI-102 exam
  • AI engineers moving from prototypes to production
  • Solution architects designing agent-based systems

Prerequisites

  • Programming experience in Python or C#
  • Familiarity with REST APIs and JSON
  • Basic Azure knowledge (AZ-900 or AI-901 level)
  • A laptop with VS Code and Python installed

Curriculum

4 Days · 32 Hours · 13 Modules · 10 Labs

  1. Planning an AI solution

    • Azure AI Foundry hubs, projects and resources
    • Choosing and comparing models
    • Deployment types, quotas and regions
    • Responsible AI planning
    • Lab: Create a Foundry project and compare two models · 40 min

    Generative AI from code

    • SDK authentication with Entra ID
    • System prompts and prompt templates
    • Structured outputs
    • Function calling
    • Streaming and token management
    • Lab: Build a chat client with function calling · 60 min

    Customising models

    • Prompt engineering vs RAG vs fine-tuning
    • When fine-tuning is justified
    • Small language models and cost trade-offs

  2. Preparing knowledge

    • Chunking strategies
    • Embeddings and vector dimensions
    • Indexing pipelines and skillsets
    • Keeping indexes current
    • Lab: Index a document set with integrated vectorisation · 50 min

    Search and retrieval

    • Vector, keyword and hybrid search
    • Semantic ranking
    • Filters and security trimming
    • Query rewriting
    • Lab: Tune retrieval quality with hybrid search and ranking · 45 min

    Grounded generation

    • Prompting with retrieved context
    • Citations and source attribution
    • Handling “I don’t know” cases
    • Lab: Build a grounded assistant with citations · 60 min

  3. Agent design

    • Instructions, tools and knowledge
    • Foundry Agent Service
    • Threads, runs and state
    • Choosing an agent framework
    • Lab: Create an agent with file search and a code interpreter · 50 min

    Tools and integration

    • Custom function and OpenAPI tools
    • Connecting to enterprise data
    • Model Context Protocol overview
    • Authentication for tools
    • Lab: Give an agent a custom API tool with scoped permissions · 55 min

    Multi-agent orchestration

    • Orchestration patterns
    • Handoffs between agents
    • Human-in-the-loop approvals
    • Failure handling
    • Lab: Orchestrate two agents with a human approval step · 45 min

  4. Vision, language and documents

    • Image analysis and multimodal prompts
    • Language services for extraction and classification
    • Document intelligence for forms and invoices
    • Speech in AI applications
    • Lab: Extract invoice data and pass it to an agent · 45 min

    Evaluation and safety

    • Evaluation datasets and metrics
    • Groundedness, relevance and safety evaluators
    • Content filters and prompt shields
    • Red-teaming basics
    • Lab: Run an evaluation and compare two prompt versions · 45 min

    Operating AI applications

    • Tracing with OpenTelemetry
    • Monitoring usage and cost
    • Deployment and versioning

    Exam readiness

    • Exam format and code-based questions
    • Practice questions with review
Download full syllabus (PDF)

Skills you will gain

  • Azure AI Foundry
  • Prompt engineering
  • Function calling
  • Embeddings
  • Azure AI Search
  • RAG
  • AI agents
  • Multi-agent orchestration
  • Document intelligence
  • Azure AI Vision
  • AI evaluation
  • Content safety
  • Tracing

Tools & technologies

  • Azure AI Foundry
  • Azure AI Search
  • Python
  • Visual Studio Code
  • Azure AI Foundry SDK
  • Azure CLI
  • Git

Certification exam

Exam details
Exam nameAzure AI Apps and Agents Developer (AI-103)
Exam codeAI-103
Voucher includedNo — book directly with the vendor
DeliveryOnline proctored or Test centre
Validity & renewal1 year (free online renewal assessment)

Skill areas measured

  1. Plan and manage an Azure AI solution
  2. Implement generative AI solutions
  3. Implement agentic solutions
  4. Implement knowledge mining and retrieval solutions
  5. Implement vision, language and document solutions

Exam details are pending verification. Vendors change exams; confirm on the official vendor page before booking. Official exam page (opens in a new tab)

Schedule & Pricing

Live Online

Individual

Price on request

  • 32 hours of live, trainer-led sessions
  • 10 hands-on labs
  • Ace Skills course completion certificate

Onsite

At your office or our centre

Pricing depends on location, group size and dates.

  • Delivered at your premises or an Ace centre
  • Same curriculum and labs as live online
  • Dates planned around your team

Corporate / Team

Private batch for your organisation

A private batch with content adapted to your environment and goals.

  • Customised curriculum
  • Private batch, online or onsite
  • Progress and assessment reports
  • GST invoice and PO-based billing

No public batches scheduled — Talk to an Advisor for a private batch.

Projects & labs you will build

  • Day 2 · 60 min

    Build a grounded assistant with citations

  • Day 3 · 55 min

    Give an agent a custom API tool with scoped permissions

Your certificate

Certificate of Completion — Ace Skills Consulting

Awarded on completing the course. It records the course title and training hours, and you can share it on your LinkedIn profile.

The Microsoft certification itself is awarded by Microsoft when you pass exam AI-103.

Frequently asked questions

I was preparing for AI-102. Is this the right course?

Yes. AI-102 retired on 30 June 2026 and AI-103 is the reported successor, with a stronger focus on generative AI and agents.

Which language do the labs use?

Python is the main language, with C# alternatives for the core exercises.

How is this different from AI-200?

AI-103 focuses on building AI apps and agents, while AI-200 focuses on broader Azure application development that integrates AI features.

What happens if Microsoft updates the AI-103 exam?

We map the curriculum to the current published skills outline and revise it when Microsoft changes the exam. Always check the latest outline on Microsoft Learn before booking.

Do you run weekend batches?

Yes. Weekend batches cover the same 32 hours across consecutive Saturdays and Sundays, with timings in IST that also suit learners in the Gulf.

Can you run a private batch for our team?

Yes. We run private online or onsite batches for corporate teams, and we can weight the labs towards your own AI use cases and data. Request a corporate quote with your team size and preferred dates.

Related courses

Explore more

All Microsoft Azure courses Microsoft certifications

Microsoft, Azure, Microsoft 365, Power BI and Microsoft Entra are trademarks of the Microsoft group of companies.

Train your team on AI-103

Private batches online or at your office in India and the GCC, with a curriculum adapted to your environment.