Why This Matters
Welcome. Over the next five weeks you are going to earn a real Microsoft certification — one that employers around the world recognise, that goes on your CV, and that opens doors. But before you can build AI on the cloud, you have to actually understand what the cloud is. Most people use the word every day and could not explain it if you asked them. By the end of today you will be able to, clearly, in your own words. That is the difference between someone who has heard of technology and someone who works in it. Let’s begin.
- “The cloud” is not mysterious and it is not in the sky. It is someone else’s computers, in a building, that you rent over the internet. That is genuinely all it is.
- Microsoft owns enormous buildings full of servers — called data centres — all over the world. When you “use Azure”, you are renting a slice of those machines.
- Why rent instead of buy? Because buying a server costs money up front, needs power, cooling, security and someone to fix it. Renting means you pay only for what you use, and you can start in minutes with nothing but a laptop.
- This is the single most important shift in computing in the last twenty years, and it is why the job you are training for exists at all.
- Scale. If ten people use your app, you rent a little. If ten million people use it tomorrow, you rent a lot — instantly, without buying anything. No individual could ever do that alone.
- Elasticity. You scale back down when demand drops, and stop paying. This is why a small startup in Lagos can run on the same infrastructure as a bank in London.
- Managed services. This is the one that matters most for you. Microsoft has already built the hard things — the AI models, the speech recognition, the vision systems — and lets you call them with a few lines of code. You do not have to invent AI. You have to learn to use it.
- That last point is the whole reason AI-901 exists as a certification, and the reason it is worth something to an employer.
- Azure is Microsoft’s cloud platform. Its main competitors are Amazon’s AWS and Google Cloud. All three do broadly similar things; Azure is heavily used by large companies and governments, which is where a lot of the jobs are.
- Inside Azure there are hundreds of services. You do not need to know them all — nobody does. You need to know the AI ones.
- The one you will live inside is Microsoft Foundry — the place where you pick an AI model, deploy it, test it, and connect your own code to it. Over half of your exam is about Foundry.
- Write this down: Cloud → Azure → Microsoft Foundry → the AI model you deploy. That is the chain you are learning. Everything else hangs off it.
- Close this page. Out loud, explain to yourself what the cloud is, without using the word “cloud”. If you cannot, read the first section again. This is not a test — it is how learning actually works.
- Now do your assignment below. Write it in your own words. I will know immediately if you have copied it, and copied words teach you nothing.
- One habit to start today: keep a single notebook — paper or a document — called AI-901 Notes. Every day, write down the one thing you did not know that morning. In 25 days you will have something valuable.
Cloud Computing
Renting computing power, storage and services over the internet instead of owning the hardware. You pay for what you use.
Data Centre
A physical building filled with servers. When you use Azure, your work is running on machines inside one of these, somewhere in the world.
Azure
Microsoft’s cloud platform. The environment your entire certification is based on.
Microsoft Foundry
The part of Azure where you find, deploy and use AI models. Over half your exam lives here. Learn this name well.
Today’s Assignment
- Roughly 200–300 words. No copying from the internet, and no asking ChatGPT to write it for you — I want your understanding, in your voice.
- It must cover: what the cloud actually is, why a business would rent instead of buy, and what Microsoft Foundry is for.
- Finish with one sentence: the thing you found most surprising today.
- Paste it straight into the submission form — there is nothing to attach and nothing to email.
Your work is marked against the published rubric and comes back to you as a written letter, usually within minutes. It is also saved to your progress page, so you can re-read every letter later. Do not submit until you have passed the self-check below.
- A. Managed services
- B. Scalability
- C. Data residency
- D. Open source
- A. A physical data centre in Nigeria
- B. A programming language used for AI
- C. The Azure environment where you deploy and work with AI models
- D. Microsoft’s competitor to Amazon AWS
- A. True
- B. False
Why This Matters
Today is a doing day, and it is one of the most important in the programme. By the end of it you will have a place of your own to run code, and — more importantly — you will understand the security model you are standing in. Every session after this one assumes you have somewhere to work. This is the only session where setup happens, so give it your full attention. Use the switcher at the top of the page to show the steps for your track. The concepts are identical either way; only the clicking differs, and the examination tests the concepts.
Go to azure.microsoft.com/free and choose Start free. Sign in with a Microsoft account, or create one against the email address you already use — that costs nothing.
Microsoft will ask for a phone number and a payment card. Read this carefully, because it is the part people panic about: the card is for identity verification, not for billing. Microsoft places a small temporary authorisation and reverses it. You are not charged, and you cannot be charged, unless you deliberately upgrade to a paid plan later. A free account gives you $200 of credit for 30 days plus a set of services that stay free beyond that. This course uses a small fraction of the credit.
If you do not have a card, or you would rather not give one, stop here and switch to Track B at the top of the page. You lose nothing that is examined.
You need no card and no Azure account. Go to aistudio.google.com, sign in with a Google account, and choose Get API key → Create API key. That is your model endpoint. It is free, it does not expire, and nobody asked you for a payment method.
Copy the key somewhere safe for now — on Day 4 you will move it into Colab Secrets, which is where it should live permanently. Treat it exactly as you would a bank card number. A key is a bearer credential: whoever holds it can spend against the account that issued it. That is true of every provider, and it is why the habit matters more than the vendor.
Then open the simulated Foundry portal from the Resources section of the course home. It mirrors the screens Track A students see, and the sessions will tell you when to use it.
Go to portal.azure.com and sign in. In the search bar at the top, type Resource groups and open it. Choose + Create.
- Subscription: your free subscription — there will only be one.
- Resource group name:
rg-yourname-ai901. Use your own name, lower case, no spaces. Mine would berg-wingo-ai901. - Region:
East US. Not every AI service is available in every region, and East US has the widest coverage. Pick it and stop thinking about it.
Click Review + create, then Create. A resource group is free. It is a folder, not a machine — it costs nothing until you put something in it.
Your equivalent of a resource group is your AI Studio project, and you already have one. Open aistudio.google.com and look at the top of the page: your key belongs to a project, and everything you create sits inside it.
Now open the simulated Foundry portal and find the Resource groups screen. Create one called rg-yourname-ai901, region East US. Nothing is really provisioned — but the navigation, the naming and the shape of the screen are the ones the examination describes, and you will be asked about them.
Write the name down in your AI-901 Notes. Later sessions will ask you where your work lives, and the honest answer for you is "an AI Studio project, with the simulator standing in for the portal." Say that plainly in your assignments. Never imply you deployed something on Azure that you did not.
rg-yourname-ai901 tells anyone who finds it what it is for and who owns it. Cloud accounts get messy fast, and tidy names are how professionals stay sane.Here are two words that get confused constantly, and the distinction is worth marks:
- A role says what you may do. Reader may look. Contributor may create, change and delete things. Owner may do all that and hand out access to other people.
- A scope says where you may do it. The same role means something very different at subscription level than on a single resource group.
Put them together and you get least privilege: give someone exactly the permissions they need, and no wider than the place they need them. A contractor brought in to build one thing gets Contributor on one resource group — not Owner on the whole subscription. This is not paranoia. It is how you limit the damage when, not if, something goes wrong.
Open your resource group and click Access control (IAM) in the left menu, then the Role assignments tab. You will see yourself listed as Owner — because you created the subscription, you own everything in it.
Now look at the Roles tab and read the descriptions of Reader, Contributor and Owner. Notice that Contributor explicitly cannot assign roles. That single line is the difference between Contributor and Owner, and it is examined.
Google AI Studio does not expose role-based access control, so read this part rather than click it — and read it twice, because it is examined and you cannot practise it.
Open the simulated Foundry portal and find Access control (IAM) on your resource group. It shows the same three roles with the same descriptions. Pay particular attention to one line: Contributor cannot assign roles. That is the difference between Contributor and Owner, and it is the single most commonly examined fact in this area.
This one is the same for both tracks, and it is not optional.
Go to learn.microsoft.com, sign in, and complete your profile. It is free and it takes two minutes. This is where your certification will live when you pass AI-901 — in your name, on your own profile, permanently, and independent of this course or anybody else’s account.
Set the legal name on that profile to match the identity document you will bring to the exam, exactly. People are turned away at the test centre over a missing middle name. Fix it now, when it costs you nothing.
Resource group
A named container that holds related cloud resources so they can be managed, permissioned and deleted together. It is a folder, and it is free.
Role
What you are allowed to do. Reader looks; Contributor builds and deletes; Owner does both and can also grant access to others.
Scope
Where a role applies — a subscription, a resource group, or a single resource. The same role means very different things at different scopes.
Least privilege
Granting exactly the access needed and no more, at the narrowest scope that works. It limits the blast radius when something goes wrong.
Today’s Assignment
- Part 1 — evidence. TRACK A A screenshot of the Azure Portal showing your own resource group, signed in as you, with the name and region visible. TRACK B A screenshot of your AI Studio project with the key page open — with the key itself blurred or cropped out — plus a screenshot of the resource group you created in the simulator.
- Also send a screenshot of your Microsoft Learn profile showing your name.
- Part 2 — understanding. In your own words: (a) What is a resource group? (b) What is the difference between a role and a scope? (c) A contractor needs to build one application for a company. What role, at what scope, would you give them, and why not more?
- Part 3 — find the edges. Look for the limits of your own access. Track A: find your billing page, then ask yourself who could see it if you had been given Contributor on one resource group instead of owning the subscription. Track B: find something in the simulator that you cannot do. Tell me what you found. Understanding the edges of access is the lesson.
- Part 4 — state your track. One line: which track you are on and where your work lives. Be accurate. Never imply you deployed on Azure if you did not — that is the one thing that would cost you a reference from me.
- If you get stuck, say exactly where. Being stuck is normal. Staying stuck quietly is the only real mistake.
Your work is marked against the published rubric and comes back to you as a written letter, usually within minutes. It is also saved to your progress page, so you can re-read every letter later. Do not submit until you have passed the self-check below.
- A. Create AI resources inside that resource group
- B. View the subscription’s billing information
- C. Grant your friend access to the same resource group
- D. Create resources in a different resource group
- A. Giving everyone administrator access so nobody is blocked
- B. Giving a person exactly the access they need for their job, and no more
- C. Removing all access until someone complains
- D. Only allowing access during working hours
- A. They mean the same thing
- B. A role is the person; a scope is the password
- C. A role says WHAT you can do; a scope says WHERE you can do it
- D. A role is temporary; a scope is permanent
rg-yourname-ai901 (scope) = full power to build, but only inside that one box. Learn it now and RBAC will never confuse you.Why This Matters
You hear “AI” everywhere now, and almost nobody using the word can define it. Today you get ahead of them. This is the day the vocabulary of your new field starts to become yours — and vocabulary is power, because you cannot think clearly about something you cannot name. Everything today maps directly onto Domain 1 of your exam, which is 40–45% of the marks. Read it slowly. This is the map of the whole territory.
- Artificial Intelligence is the big outer box: software that does things we would call intelligent — recognising a face, understanding a sentence, making a prediction.
- Machine Learning sits inside it: instead of a human writing the rules, the system learns the rules by looking at huge amounts of examples. Show it a million photos labelled ‘cat’ and it learns what a cat looks like. Nobody wrote a rule that said “cats have whiskers”.
- Generative AI sits inside that: models that do not just recognise or predict, but create — text, images, code, audio. ChatGPT is generative AI. So is an image generator.
- Every one of those is on your exam. Get the nesting right: AI > Machine Learning > Generative AI.
- Generative and agentic AI — creating content, and AI ‘agents’ that can take actions and use tools to complete a task, not just answer a question.
- Text analysis — reading language and pulling meaning out of it: sentiment analysis (is this review angry or happy?), entity detection (which words are names, places, dates?), key phrase extraction (what is this document about?), summarisation (make it shorter).
- Speech — speech recognition turns spoken audio into text; speech synthesis turns text into a spoken voice. Two directions, two names. Know both.
- Computer vision — understanding images: what objects are in this photo, is there text in it, whose face is this. Plus image generation, which creates new images.
- Information extraction — pulling structured data out of messy sources: invoices, forms, receipts, images, audio, video. Enormously valuable in real businesses, and it is a whole section of your exam.
- Learn these five names. The exam will describe a scenario and ask you which workload it is. That is a free mark if you know the list.
- A bank scans a customer’s ID document and pulls out the name and date of birth automatically — that is information extraction.
- A shop reads a thousand customer reviews and finds out how many are negative — text analysis (sentiment).
- A voice note gets turned into text you can read — speech recognition.
- A support chatbot that can look up your order and issue a refund on your behalf — agentic AI. Note the difference: it does not just reply, it acts.
- This is the work. These are the problems companies pay people to solve. You are learning to be the person who solves them.
- Say the five workloads out loud from memory: generative and agentic, text analysis, speech, computer vision, information extraction. Repeat until you do not have to look.
- Write in your AI-901 Notes the four text-analysis techniques: sentiment analysis, entity detection, key phrase extraction, summarisation. These four come up again and again.
- Then do the assignment. Take it seriously — this one genuinely tests whether you understood.
Machine Learning
Software that learns patterns from large numbers of examples, rather than being given explicit rules by a programmer.
Generative AI
AI that creates new content — text, images, code, audio — rather than only classifying or predicting.
Agentic AI
AI that can take actions and use tools to accomplish a goal, not just produce an answer. A major focus of the AI-901 exam.
Sentiment Analysis
A text-analysis technique that judges whether language is positive, negative or neutral. One of four you must know.
Today’s Assignment
- Write out each scenario and answer it:
- (1) An app listens to a doctor speaking and produces written notes. (2) A company scans 5,000 paper invoices and pulls the total amount from each. (3) A tool writes a first draft of a marketing email. (4) A security camera identifies whether a person is wearing a hard hat. (5) A system reads 10,000 tweets about a product and reports how many are angry.
- For each: name the workload, and write one sentence on why. The ‘why’ is what I am marking.
- Then, from your own life in Nigeria, give me one example of a problem you think AI could solve. One paragraph. This is the question I most want you to think about.
Your work is marked against the published rubric and comes back to you as a written letter, usually within minutes. It is also saved to your progress page, so you can re-read every letter later. Do not submit until you have passed the self-check below.
- A. Computer vision
- B. Text analysis (sentiment analysis)
- C. Speech synthesis
- D. Image generation
- A. It runs faster
- B. It is trained on more data
- C. It can take actions and use tools to complete a task, not just produce an answer
- D. It only works with images
- A. Speech recognition
- B. Speech synthesis
- C. Entity detection
- D. Summarisation
Why This Matters
Today you write code for the first time. Microsoft’s own guidance for this exam says you need to know Python syntax — so we are not going to pretend otherwise, and we are not going to leave you to discover that in Week 3 with no preparation. You do not need to become a software engineer. You need to be able to read a short Python script, understand what it is doing, and change it without fear. That is entirely achievable in the time we have. Expect to feel clumsy today. Everyone does. It passes.
- Open
colab.research.google.comin your browser and sign in with a Google account. This is Google Colab — it lets you write and run Python in the browser, free, with nothing to install. - Click New notebook. You will see an empty box (a ‘cell’). Type code into it, press the play button or
Shift+Enter, and it runs. - This removes the single biggest thing that makes beginners quit: fighting to install software before they have written a line of code. We skip that entirely.
- Type this, and run it:
print("Hello, my name is ")— put your own name inside the quotes. That is it. You have now written and run a program. You are a person who codes.
- Variables — a name that holds a value.
name = "Ada"andage = 19. The=means ‘store this’, not ‘is equal to’. - Types — values have kinds. Text is a string (in quotes:
"hello"). Whole numbers are integers (19). Decimals are floats (19.5). True/false values are booleans (True,False). Mixing these up is the number one beginner error. - print() — shows you a value. Your window into what the machine is thinking. Use it constantly.
- Comments — anything after a
#is ignored by Python. It is a note to a human. Write them. Your future self will thank you. - Try this in Colab, line by line, using your own name:
name = "YOUR NAME"goal = "pass AI-901"days_left = 25print(name + " will " + goal + " in " + str(days_left) + " days") - That
str()is important: you cannot glue a number directly onto text. You must convert it to a string first. If you get an error there, good — read it. Errors are instructions, not insults.
- Code becomes useful the moment it can choose. That is what
ifdoes. - Type this exactly, indentation and all:
score = 750if score >= 700:print("You passed AI-901!")else:print("Not yet. Go again.") - Note two things. The colon at the end of the
ifline. And the indentation — those four spaces are not decoration. In Python, indentation is how the language knows what belongs inside theif. Get it wrong and it breaks. - Now change
scoreto650and run it again. Watch the output change. You made that happen. - By the way — 700 really is the AI-901 pass mark. Keep that number in your head.
- Deliberately break something. Remove a quotation mark and run it. Read the error message. Errors are how Python talks to you — learning to read them calmly is half of learning to code.
- Save your notebook. In Colab: File → Save a copy in Drive. Name it AI901-Day4.
- Do the assignment below. Type it yourself, run it until it works, and send me the result — including any errors you fought through. I am as interested in the fight as the finish.
Variable
A name that stores a value, so you can use it later. name = "Ada"
String / Integer
A string is text, written in quotes. An integer is a whole number. Confusing the two causes most early errors.
Indentation
The spaces at the start of a line. In Python this is not style — it is meaning. It defines what code sits inside an if, a loop, or a function.
Google Colab
A free browser tool for writing and running Python with nothing to install. Your coding workspace for the next three weeks.
Today’s Assignment
- In Google Colab, write a program that: stores your name, your target exam score (700), and a variable called
practice_score. - Use an
if / elseto print “[Your name] is ready for AI-901” ifpractice_scoreis 700 or above, and “[Your name] needs more practice — keep going” if it is below. - Run it twice — once with a passing score and once with a failing score — so you see both messages.
- Add at least two
#comments explaining what your code does. - Send me: the Colab share link (or a screenshot of the code and its output), plus one sentence on the error that gave you the most trouble and how you fixed it. If you had no errors, you are not typing it yourself.
Your work is marked against the published rubric and comes back to you as a written letter, usually within minutes. It is also saved to your progress page, so you can re-read every letter later. Do not submit until you have passed the self-check below.
print(type(19.5)) tell you the value is?- A. An integer
- B. A string
- C. A float
- D. A boolean
19 would be an integer, "19" would be a string, and True would be a boolean.print("Days left: " + 25) cause an error?- A. Because 25 is too small a number
- B. Because you cannot join a string and an integer directly — the number must be converted with str()
- C. Because print() only accepts one value
- D. Because the quotation marks are wrong
"Days left: " + str(25). This is one of the most common errors beginners meet — and now you already know it.- A. Nothing — it just looks tidy
- B. It makes the code run faster
- C. It defines which lines belong inside a block, such as an if statement
- D. It is only needed in comments
Why This Matters
Last day of the foundation week. Today you learn the final pieces of Python you actually need — and then you see, for the first time, the shape of the code you will be writing for the rest of this programme. Every AI application you build in Weeks 3, 4 and 5 follows the same simple pattern, and by the end of today you will recognise it. Then take the weekend. You have earned it.
- A list holds many values in order:
workloads = ["vision", "speech", "text"] - Get one out by its position, counting from zero:
workloads[0]gives"vision". Yes, zero. Programmers count from zero. You will get used to it. - A loop repeats an action for every item:
for w in workloads:print("I am learning: " + w) - Note the colon and the indentation again — the same grammar as
if. Python is consistent, which is a kindness. - That loop just did three things with three lines. With a list of ten thousand customer reviews, it would do ten thousand things with the same three lines. That is why this matters.
- A function is a named block of code you can run whenever you want, as many times as you want.
def check_score(score):if score >= 700:return "Pass"else:return "Not yet"- Then use it:
print(check_score(750))andprint(check_score(600)). Write the logic once, use it forever. defdefines it. The word in brackets is what you give it.returnis what it hands back.- Almost every piece of AI code you will write is: call a function, give it some input, get a result back. If you understand this section, you understand the shape of everything ahead.
- You will not write AI models. You will call them. Here is the pattern — read it, do not worry about running it yet:
# 1. Import the tool you needfrom azure.ai.something import Client# 2. Connect, using your secret keyclient = Client(endpoint, key)# 3. Send your input, get a resultresult = client.analyze("This product is terrible")# 4. Use the resultprint(result.sentiment)- Four steps. Import. Connect. Send. Use. That is it. That is the whole shape.
- An endpoint is the web address of the AI service. A key is your password to it. You will get both from Foundry, inside your own resource group.
- And remember Day 2: never post a key anywhere public. It authorises spending against a real subscription — mine. Treat it like a bank PIN. Careless engineers leak their keys; careful ones get hired.
- Every single application in the rest of this course is a variation on those four lines. When Week 3 arrives and it looks intimidating, come back and read this again. You already know the shape.
- Look back at Monday. Five days ago you could not define the cloud. Today you have written a function, you are inside a real Azure subscription, and you understand how an AI application is structured. That is real progress and you should let yourself feel it.
- Next week you begin the exam syllabus properly — Domain 1, responsible AI, and how generative models actually work.
- Do the assignment, send it in, then rest. Consistency beats intensity. I would far rather have 30 focused minutes every day than four hours on a Sunday.
List
An ordered collection of values. Indexed from zero: the first item is [0].
Loop (for)
Repeats an action for every item in a list. How you process a thousand things with three lines of code.
Function
A named, reusable block of code. Defined with def, hands a value back with return.
Endpoint & Key
The address of an AI service, and the secret that proves you may use it. Both come from Foundry. Never share the key.
Today’s Assignment
- Part 1 (code). In Colab, create a list of the five AI workloads you learned on Day 3. Write a function called
describe(workload)that returns a sentence about it. Then use aforloop to print a description of all five. - Part 2 (understanding). In your own words, write out the four steps that every AI application in this course will follow, and explain what an endpoint and a key are — and why the key must never be shared.
- Send me the Colab link (or screenshots of code and output) plus your Part 2 answer.
- And one more thing. Tell me honestly: what was the hardest part of this first week, and are you finding 30 minutes a day workable? I need to know so I can pitch next week correctly. An honest answer is not a complaint — it is information I need.
Your work is marked against the published rubric and comes back to you as a written letter, usually within minutes. It is also saved to your progress page, so you can re-read every letter later. Do not submit until you have passed the self-check below.
workloads = ["vision", "speech", "text"], what does workloads[1] return?- A. "vision"
- B. "speech"
- C. "text"
- D. An error
[0] is "vision" and [1] is "speech". This trips up almost every beginner exactly once. Now it has tripped you here rather than in an exam.return do?- A. Prints a value to the screen
- B. Restarts the program
- C. Hands a value back to whatever called the function
- D. Ends the whole script
return and print are different and beginners confuse them constantly. print shows a human. return gives the value back to your code so it can be used.- A. Paste it into a public GitHub repo so you do not lose it
- B. Share it in a forum when asking for help
- C. Keep it secret — it authorises real spending on a real subscription
- D. It does not matter, it is only a test account