Historical AI-900: Azure AI fundamentals

Objectives and progression

Independent course with five modules and 32 original decisions in fictional APS and IT-project contexts. Internal assessment of 20 decisions in 45 minutes. Distinguishes historical capabilities from current service lifecycle. The AI-900 exam retired on 2026-06-30. This course retains the 2025-05-02 objectives for historical learning. Azure AI Fundamentals continues with AI-901, whose syllabus differs. This is not complete AI-901 preparation.

Audience: IT professionals, support teams, and managers evaluating AI applications.

Prerequisites: Cloud and client-server application basics. These historical objectives require no prior data science experience.

220 estimated study minutes

• Choose the AI task and define acceptance conditions and human control.
• Distinguish tasks, training data, and evidence of generalization.
• Match visual output to the requirement and validate extracted data.
• Choose the right transformation and preserve meaning in integrations.
• Evaluate generation, retrieved context, and limits before using recommendations.

Modules

1. Workloads and operational responsibility
2. Machine learning and useful evaluation
3. Vision and documents with validation
4. Language, speech, and operational meaning
5. Generative AI and grounded responses

Continue learning

Data / AI

• AZ-900 — Azure Fundamentals
• DP-900: Azure Data Fundamentals
• AWS Certified AI Practitioner

References and version

AI-900 historical skills measured 2025-05-02; exam retired 2026-06-30

• AI-900 historical study guide and retirement · 2026-09-30
• AI-901 current study guide · 2026-09-30
• Azure AI Fundamentals current credential · 2026-09-30
• Responsible AI principles · 2026-09-30
• Automated machine learning · 2026-09-30
• Machine learning task types · 2026-09-30
• Overfitting imbalance and target leakage · 2026-09-30
• Model management deployment and lineage · 2026-09-30
• Vision capabilities and Image Analysis 4.0 lifecycle · 2026-09-30
• Face detection and recognition · 2026-09-30
• Document Intelligence extraction and layout · 2026-09-30
• Extraction confidence and evaluation · 2026-09-30
• Language workloads and legacy capabilities · 2026-09-30
• Speech transcription synthesis and translation · 2026-09-30
• Model catalog capability and availability · 2026-09-30
• Retrieval augmented generation · 2026-09-30
• Direct and indirect prompt attacks · 2026-09-30
• Tokens embeddings and transformer attention · 2026-09-30
• Model benchmarks and their scope · 2026-09-30
• Microsoft Foundry platform and historical naming · 2026-09-30
• Evaluation of generative applications · 2026-09-30
• Technical exam scaled scoring · 2026-09-30