
AI-103 Azure AI Apps and Agents: Exam Skills Explained
About this course
This course contains the use of artificial intelligence. AI-103 is the exam for the Azure AI Apps and Agents Developer Associate certification, and its skills outline is wide: setting up Microsoft Foundry, generative AI and agents, images and video, text and speech, and search and document extraction. This course explains that outline in plain English, one skill at a time, and follows a single example company from its first model call to a release that moves through test and production.
It is an explain-and-practice course. Each idea is placed on a map, tied to the example, and then checked with short scenario exercises that come with model answers. The example.
Larkfield Mutual is a made-up home and travel insurer. It is a composite case, so no real company sits behind it, and its numbers (about 1,800 new claims a month, six claims handlers, 400 policy PDFs) are invented to feel realistic. Its claims assistant is built on Microsoft Foundry one capability at a time, and one claim, C-1042, brings an email, a repair quote, two photos, a short clip and a phone call through the whole build.
Every story in module 2 uses this company, so no idea floats in the air. What makes the course different. Plain words first.
Module 1 opens with a key-terms lesson, and every new word comes with a thing you already know and a note on what it is confused with. One example from start to finish, instead of a new toy case in every lesson. A map, not a list.
Each module opens with a map, each lesson says which step of the map it covers, and each of the 64 skill statements of the study guide has a lesson that explains it and a lesson that applies it. Honest limits. The lessons say where things break, what is still preview, and what the course cannot show, and this page says what the course leaves out.
The three modules. Foundry basics and what AI-103 checks (14 lessons, 1 h 47 min): the six starter words, what an Azure AI developer builds and what the exam tests, a plain model call versus a search app versus an agent, choosing between large, small, reasoning and multimodal models, how a Foundry resource, project and deployment fit, deployment types, quotas and rate limits, tools, memory and workflows, embeddings and hybrid search, guardrails and evaluators, images and video, text and speech, and document reading. It ends with a short scenario exercise that picks a model, a deployment type and a search method for three jobs.
Larkfield Mutual builds its claims assistant (16 lessons, 2 h 26 min): thirteen stories, from the first model call and a claim email turned into structured JSON, through policy PDFs in a search index, grounded answers with retrieval-augmented generation, repair quotes read with Content Understanding, damage photos and clips understood, images and short videos made and edited, a claim call transcribed and analyzed, a triage flow of two models, a claims agent with tools, a second agent with a human approval step, guardrails against unsafe text and tricked images, to evaluation, tracing and release. Two scenario exercises close the module: writing a tool schema with an approval rule, and picking a search type and chunk size for three questions. How it lives at work and what to do next, the last module (11 lessons, 1 h 21 min): sizing quotas, scaling and cost, securing Foundry with identities, roles and networks, governing agents with oversight, limits and audit, evaluating and red-teaming an app, monitoring drift, safety, index health and traces, reading common errors, moving a project through CI/CD, tuning prompts and adding rules beside the model, a scenario exercise that traces three failures to their cause, and a closing lesson on how the exam is built and what to study next.
This module comes with its memo (99_memo. pdf), a short quiz and its practice sheet (99_practice. pdf).
Every module ends with a two-to-four page memo as a PDF, a short optional quiz, and a practice sheet with the exercise tasks first and the model answers at the back. What this course is not. It is not lab work in Azure.
There is no screen recording of the Foundry portal and no click-through, and you do not need an Azure subscription. Code on the slides follows patterns from the documentation, is shown as an example and was not run against Azure, and the outputs of Azure services that appear in the lessons are labeled as illustrative. It is not a practice-test bank.
The exercises are short scenario choices, a few per module, and there are no full practice exams here. It is not a promise about exam results. Microsoft publishes the outline and the exam rules, and the course explains the skills the outline lists.
Nobody can promise a score, and this course does not. It does not teach Python from zero, and it does not replace building an app in your own subscription, which you can do after it. It is independent training from InSteppe Academy.
It is not affiliated with or endorsed by Microsoft. The names AI-103, Azure and Microsoft Foundry are used only to describe the subject. How the outline is covered.
The lesson plan follows the five skill areas of the study guide: plan and manage an Azure AI solution, implement generative AI and agentic solutions, implement computer vision solutions, implement text analysis solutions, and implement information extraction solutions. Each skill statement is explained once, in module 1 or module 3, and applied once, in the Larkfield stories or in a scenario exercise. The time given to each area follows its weight in the outline within a few points.
Depth differs by topic, and the exam may ask about related topics beyond what the lessons show. Use the course as a map, and practice in a real subscription after it, if you have one. The video runs 5.
65 hours (5 h 39 min) in 42 lectures, shorter than the typical course for this exam on purpose: it explains the decisions and leaves out screen tours.
Skills you'll gain
Available Coupons
Course Information
Level: Intermediate
Suitable for learners at this level
Duration: Self-paced
Total course content
Instructor: InSteppe Academy
Expert course creator
This course includes:
- πΉVideo lectures
- πDownloadable resources
- π±Mobile & desktop access
- πCertificate of completion
- βΎοΈLifetime access
You May Also Like
Explore more courses similar to this one


