Every marketing course in Cairo has added an AI module in the last eighteen months. Most of them teach prompts. A prompt list is worth about a week before the model changes, and it does not survive the first time a client asks why the automation spent the budget the way it did.
So here is what an AI module should contain if it is going to be worth paying for, and the test you can run on any syllabus before you enrol. The short version: a useful module teaches you to read what the platform documents about its own automation, because that is the skill that keeps working after the tools change.
The thing that makes AI modules go stale
The tools are not the subject. The subject is that the ad platforms have moved decisions from you to their models, and your job changed from operating controls to setting constraints and reading results. A course built on tool tours teaches the surface of that. A course built on the documented behaviour of the automation teaches the thing underneath, which is why it is still true a year later.
That distinction also tells you what to do with a course that is just prompts. It is not useless, it is just a one week topic charging for twelve.
Read the vendor's own number, including its footnote
Here is the exercise I would want in week one of any AI module, using a live example. Google's help page for AI Max for Search campaigns states that advertisers "will typically see 14% more conversions or conversion value at a similar CPA/ROAS".
Now read the footnote, because this is the actual lesson. Google attributes that figure to its own internal data from 2025, and the page says it applies to non-Retail advertisers. So the number is first party, not independent, it describes a typical case rather than a guaranteed one, and it explicitly excludes retail, which is a large share of the clients an Egyptian agency actually serves.
None of that makes the figure dishonest. Google published the qualification itself. The point is that a marketer who quotes "14% more conversions" to a client without the qualification has made a claim the documentation does not support, and a course that does not teach this distinction is teaching you to repeat vendor marketing.
This is the same discipline that a good analytics module teaches about any platform number, which is why the two modules belong next to each other. Our analytics course outline covers the measurement half of it.
Know what the automation overrides
The second skill is harder and more valuable: knowing which of your settings the automation still respects and which it quietly overrides. Vague answers here cost real money.
Google's AI Max FAQ is unusually specific, and three lines from it are worth memorising:
- "Negative keywords will be respected even with AI Max turned on." Your exclusions survive.
- If both text customization and Final URL expansion are on, pinned assets in responsive search ads are not respected. Pinning is also ignored when URL inclusions are provided. So a brand that legally requires fixed wording in a headline has a real conflict to resolve.
- Turning AI Max off disables its features including brand exclusions, and assets such as grammar and special offers are fully automated with no option to customise them.
Read that middle point again if you work with regulated clients. A pharmacy, a clinic or a financial product often has approved wording that must appear exactly. "The automation may ignore your pinned headline" is the kind of sentence that should appear in a course, and almost never does.
Learn the tooling gaps too
The same FAQ states that AI Max "is not yet available in the API and Editor", and that toggling it may cause API errors for text customization and brand settings. That matters for anyone managing accounts at scale, because bulk workflows through Editor are how agencies actually work in practice.
The teaching point is not the gap itself, which will close. It is that platform features arrive in the interface before the bulk tools, so part of evaluating any new automation is asking what it breaks in your existing process. A course that only shows you the interface will never raise that question.
What an AI module should actually cover
Taking all of that together, here is the content I would expect for the money, in order:
- Where the decisions moved. What the automated campaign types decide now, and what inputs you still control.
- Reading platform documentation. Finding the eligibility conditions, the attribution window and the qualifications on any published figure. A practical session with real help pages open.
- Generative work with a review step. Copy and creative variants produced quickly, then a defined check before anything runs. The check is the skill.
- Measurement that survives automation. How to tell incremental results from attributed ones when the system optimises across paid and organic together.
- Where it still needs a person. Claims, compliance, local dialect, and anything with legal exposure.
- What it costs. Honest per seat tool pricing, because a workflow built on four paid tools is not a workflow a junior can take to a small employer.
Notice that only one of those six is about the tools. That ratio is the test.
Four questions to ask before you pay
- Which platform documentation will we read in class? If the answer is none, the module is a tool tour.
- What do I produce by the end? A campaign with a written measurement plan beats a folder of generated copy.
- What are the tool costs after the course? You need the number to know whether you can use what you learned.
- What does the trainer say AI is bad at? A trainer with no list has not used it under pressure.
If you are earlier than this and still choosing a first course, which tools to learn first and the three month plan are the better starting points, and the full module list is in the diploma breakdown.
Honest limits, including one about us
First, the obvious one: an AI module cannot make you employable on its own. Employers in Cairo hire for the whole job, and "knows AI tools" is a line on a CV, not a role.
Second, this field moves faster than any syllabus. The specific behaviours above were accurate on the Google help pages at the time of writing, and the API gap in particular is stated as temporary. A course that promises a permanent tool stack is promising something it cannot deliver. What lasts is the habit of checking the documentation before you believe a claim, including ours.
Third, and this costs us a sale: if your job today has no ad budget and no account to practise on, an AI module will be abstract for you. Learn the fundamentals first and come back to this when you have something to automate. The same applies in reverse for engineers: if what you actually want is to build systems rather than run campaigns, a marketing course is the wrong purchase, and this write up on AI lead scoring and when the score is just a guess is closer to the work you are describing.
Fourth, we do not run a standalone AI course and we are not planning one. The material above sits inside the diploma, because teaching it apart from media buying and measurement is what produces the prompt courses this post is about.