AI-901 Tests Whether You Can Match AI Workloads to Foundry Tools and Use Them Responsibly
AI-901, which replaced the retired AI-900, rewards recognising the right AI workload, choosing and configuring a model, and building simple Foundry apps that follow responsible AI principles.
Check Your Readiness →Most candidates understand Microsoft Certified: Azure AI Fundamentals (AI-901) concepts — and still fail. This exam tests how you apply knowledge under pressure.
The first three steps map to the AI concepts and capabilities domain (40-45%) and the last two to implementing AI solutions in Microsoft Foundry (55-60%). Most items describe a scenario and ask which workload, tool, setting or principle applies.
Choose transparency because the bank should explain how the model reached its decision.
Choose fairness. A model that treats similar applicants differently because of a demographic group is a fairness problem, even though transparency is also valuable.
Send the invoice images to an image generation model or a plain chat prompt and copy the answer.
Use Azure Content Understanding in Foundry Tools to extract structured fields from documents and forms. Information extraction is its own workload with a purpose-built tool.
Raise the temperature so the model explores better answers.
Lower the temperature and give a precise system prompt, ideally with the approved FAQ content as grounding. Consistency comes from lower randomness and clear instructions.
Transparency means people can understand how an AI system works, what it is for and its limitations. Accountability means people and organisations remain responsible for the system's outcomes and its governance, so a question about who answers for a decision points to accountability.
Temperature controls randomness: lower values give more focused, repeatable output and higher values give more varied, creative output. It does not add knowledge, so it cannot by itself stop a model from producing incorrect content.
The system prompt (system message) sets the assistant's role, tone and constraints for the whole conversation, while the user prompt carries the specific request. Exam items often ask where to place an instruction such as "only answer questions about our products", and the answer is the system prompt.
Entity recognition identifies and categorises items such as people, places, organisations and dates, while key phrase extraction returns the main talking points of a text. Sentiment analysis and summarization are separate tasks again, so read what output the scenario wants.
Speech recognition (speech to text) turns spoken audio into text, and speech synthesis (text to speech) turns text into spoken audio. A voice assistant uses both, but a captioning scenario needs only recognition.
Strengthen weak areas with exam-style practice questions and detailed explanations.
Simulate the real exam experience and assess your readiness under timed conditions.
Review key concepts, objectives, and exam topics in one place.
Get personalized explanations, learning recommendations, and instant answers.
Follow a structured learning path designed to help you prepare efficiently.
See whether Foundry hands-on tasks or AI concepts are holding back your AI-901 score.