AI-901 Study Plan¶

Four weeks at 4 to 6 hours per week. One more week than most fundamentals plans, because more than half the exam is hands-on Foundry work: you should actually deploy a model, build an agent and run Content Understanding, not just read about them. An Azure free account or a pay-as-you-go subscription with a budget alert is enough; delete resources when you finish.

Week 1: Concepts and responsible AI (Domain 1)¶

Read the study guide and its change log
Learning path module: Get started with AI in Azure
The six responsible AI principles; write one scenario of your own for each
How generative models work: tokens, transformers, embeddings, multimodal, hallucination, grounding, agents
Choosing a model by capability; deployment types (Global, Data Zone, regional; standard vs provisioned)
Parameters: temperature, top-p, max tokens, system message
Workload recognition: generative, agentic, text, speech, vision, image generation, information extraction
Review: notes/01-ai-concepts-and-responsible-ai.md

Week 2: Generative AI apps and agents (hands-on)¶

Learning path module: Get started with generative AI and agents in Azure
Create a Foundry project; deploy a small chat model; test in the playground with a system message
Run the Python quickstart (azure-ai-projects, DefaultAzureCredential, responses.create)
Turn it into a chat loop
Create a prompt agent in the portal with instructions and one tool; test it
Run the agent client quickstart with a conversation and a follow-up question
Review: notes/02-genai-apps-and-agents-in-foundry.md

Week 3: Text, speech and vision (hands-on)¶

Learning path modules: text analysis, speech, computer vision
Sentiment, key phrases and entities on sample reviews (Azure Language), then the same with a prompt; compare
Transcribe a short recording; synthesise speech with SSML
Send a spoken prompt to an audio-capable model
Send an image with a question to a vision-capable model; generate an image
Review: notes/03-text-speech-and-vision.md

Week 4: Information extraction, Foundry IQ and review¶

Learning path modules: AI-powered information extraction; Microsoft Foundry IQ
Run a prebuilt Content Understanding analyzer on an invoice or receipt
Build a custom analyzer with three fields (extract, classify, generate); inspect confidence scores and grounding
Review: notes/04-information-extraction.md
Practice questions; review every wrong answer
Microsoft's free practice assessment on AI Skills Navigator; aim for 80%+ before booking

Exam week¶

Re-read the "Names you must know" table in the fact sheet
Re-read the responsible AI trigger-word table
Delete your Foundry resources