AI-901 Exam Map and Crosswalk

First, the naming trap: AI-900 retired June 30, 2026. The current Azure AI Fundamentals exam is AI-901 — same badge ("Microsoft Certified: Azure AI Fundamentals"), same $99 fee, but a meaningfully updated syllabus. Ignore any prep material that hasn't been rewritten for AI-901; it will teach you retired service names and miss the heaviest new domain.

The exam at a glance

• ~40–60 questions, 60 minutes — roughly a minute per question, which is plenty.
• Question formats: single-choice, multiple-response ("select two"), drag-and-drop matching, and hot-area screens. No live labs, no code writing — it's entirely question-based, even the "hands-on" domain.
• Passing: 700 of 1,000 (scaled — not simply 70% of questions).
• Delivery: Pearson VUE test center or online proctored from home. $99.
• No prerequisites, and the credential doesn't expire (fundamentals badges don't).

What changed from AI-900 — and why it favors you

AI-900 asked you to describe services. AI-901 asks you to understand building AI applications and agents on Microsoft Foundry. The syllabus now leads with generative AI, RAG and grounding, prompt engineering, agents, and Copilot fundamentals — with roughly 40–45% on AI concepts and 55–60% on implementation with Microsoft Foundry. Generative AI is the heaviest single skill area.

Read that list again: RAG, grounding, prompt engineering, agents. The exam moved onto your home turf. Most candidates coming from the old AI-900 material find the new content foreign; you've been drilling it for weeks.

The crosswalk — what you already own vs. what's new

You already own (skim, don't study):

• Generative AI, LLMs, prompts → Interview Prep Module 12 lesson 1, Module 13 lesson 1.
• RAG and grounding → Module 12 lessons 2–3, Module 13 lesson 2. The exam's RAG questions are simpler than what you've drilled.
• Agents and tool use → Module 12 lesson 8, Module 13 lesson 3.
• Classical ML basics (regression, classification, clustering, features/labels) → Module 12 lesson 6, Module 13 lesson 5. The exam stays at exactly this depth.
• Responsible AI concepts → Module 11 lesson 3 — but learn Microsoft's specific six principles by name (lesson 4 of this module).

Genuinely new for you (the gap lessons in this module):

1. Current Azure service names and what-service-for-what (lesson 2) — the exam's bread-and-butter trivia, and it tests the current names.
2. Microsoft Foundry essentials (lesson 3) — the platform at the center of the new exam: hubs, projects, model catalog, deployments, agent service, evaluations.
3. The Microsoft six Responsible AI principles + workload taxonomies (lesson 4) — exam-specific vocabulary for things you already understand.
4. Question tactics and the study plan (lesson 5).

Realistic effort estimate

For someone starting cold: 2–4 weeks. For you: the concept half of the exam is already banked — the work is the Foundry/service-names layer, which is this module plus the free official Microsoft Learn path and practice assessment. A focused week is credible; two is comfortable. That makes this the fastest possible route to a current, official Microsoft AI credential in hand.