AI makes you faster at whatever you already are

People keep asking this school whether we are for AI or against it. It is the wrong question, and we would rather answer the right one: what is it actually for, what is it not for, and what does a marketer need to know before touching it? This is where we stand.

Our first cohort has had four Saturdays. This week is advertising and campaigns, and an assistant will be open on somebody’s laptop before the second hour is out, because that is how the work is done now.

So let us be plain about the position. This school is not neutral on AI. We use it, we teach with it, and we think a marketer who refuses to touch it is choosing to be slower than the person they are competing with for the same job.

What we will not do is pretend it is a shortcut around knowing your work.

Start with what it is actually good at

Strip away the noise and these tools do one thing very well: they predict what should come next, from patterns in what has come before. The economist Ajay Agrawal calls it a prediction technology, and that description is more useful than most of what you will read about it, because it tells you exactly where it helps and exactly where it will let you down.

Prediction is worth a great deal in marketing. A week of work can genuinely be compressed.

Everything in the second column is judgement. It is also, not coincidentally, the part of marketing that gets people hired and keeps clients paying.

The evidence that tools alone do not pay

Nearly nine in ten organisations in McKinsey’s global survey this year use AI in at least one part of the business. That part is settled. But only 37 per cent can point to any effect on profit, and only about six per cent qualify as high performers, meaning AI accounts for a meaningful share of their earnings.

What separates that six per cent is not better tools. Everyone has the same tools. Nearly three quarters of the high performers had fundamentally redesigned how the work flows around AI. Among everyone else, about one in four had.

This is the whole reason this school teaches a five-stage system rather than a software menu. Positioning, content, distribution, capture and conversion, measurement. Put an assistant inside a system and it compounds. Put one inside a habit of posting whenever you remember to, and you get the same habit, faster.

What bad AI work costs, and who pays for it

There is a word for the output of a tool used without judgement, and it arrived last year: workslop. Work that looks finished and is hollow. Researchers at BetterUp Labs and Stanford found 41 per cent of workers had been sent it, and that each instance cost roughly two hours to sort out.

Two hours, paid by the person who received it.

In an office that is annoying. In marketing it is worse, because the person receiving it is your client, and they do not send it back for correction. They read a report full of confident numbers that do not reconcile, or a caption that could belong to any business in Kampala, and they quietly decide you are not worth the retainer. You never learn why.

That is the real risk of these tools in our market. Not that AI takes the job. That it helps somebody produce enough convincing rubbish to lose the client faster than they could have managed alone.

It rewards the people who already know things

There is a finding in this research that deserves more attention than it gets, especially from anyone starting out.

When researchers test AI assistance on writing, analysis and similar work, the people who gain most are usually the ones who were already strong. The tool steepens the slope rather than flattening it. If you know what good looks like, it gets you there faster. If you do not, it helps you produce more of what you cannot evaluate.

The economist Carl Benedikt Frey has put the long-term worry well: if organisations hire fewer juniors and train less because the machine covers the basics, they erode the very expertise needed to check the machine’s output.

Which is why our answer to a twenty-three-year-old asking whether to learn AI or learn marketing is: that is not a choice, and anyone selling it to you as one is selling you something.

We have seen this film before in Uganda

Ninety-six per cent of Ugandans live under 4G coverage. Around twenty-two per cent actually use mobile internet. The network arrived years ago; the usage did not follow, because coverage was never the thing in short supply.

The same report scores Uganda 70 out of 100 on digital government and under 10 on digital business. Our institutions went digital. Our businesses largely did not.

The tools are already here, most of them free, all of them on the same phone everyone is holding. That means the advantage will not go to whoever gets access first. It goes to whoever knows what to ask for, and can tell whether what came back is any good.

What the companies building these tools can see

The people who make these systems publish their own usage data now, and it says more than any of the advertising around them.

Microsoft measured how many working-age people actually used generative AI in the first three months of this year, country by country. The global average came to 17.8 per cent. This is our neighbourhood.

Microsoft, Global AI Diffusion, first quarter of 2026. Rwanda came last of the 46 African countries measured.

Developed countries averaged 27.5 per cent and developing countries 15.4, and over the year that distance grew rather than closed. Uganda sits at roughly one working-age person in thirteen.

Now read that as a marketer rather than as a statistic. Twelve of the thirteen people competing with you for the same job, or pitching the same business owner, are not using these tools for their work at all. You are not late. You are early, in a market that has not noticed yet.

Anthropic publishes a similar picture from its own traffic, and adds something more uncomfortable. Usage tracks income almost mechanically: roughly speaking, a one per cent rise in income per head comes with a 0.7 per cent rise in usage. When they compared the second half of last year with the months before it, they found no sign of lower-usage countries catching up.

But the finding that should interest anyone learning this work is not how much these tools are used. It is how.

In countries where usage is low, people are more likely to hand over a whole task and accept whatever comes back. Where usage is high, people work alongside the tool: asking, correcting, asking again. Same software, two completely different habits.

And OpenAI’s own study of how people use ChatGPT found something that will sound familiar to anyone teaching in this region. More than 700 million people use it weekly, about one adult in ten on earth, and the fastest growth is in low- and middle-income countries. Yet close to three quarters of those conversations have nothing to do with work.

That is exactly what we see in our classrooms. The tool is already in the hand. It is being used to word a difficult message, to help with a child’s homework, to settle an argument. The leap is not getting people to use AI. It is getting them to point it at their work, and to know what to do with what comes back.

So here is our stand

What that looks like in the classroom

None of the above is a slogan we put on a poster. It is how the eight weeks are taught.

And for people already doing this work who do not need eight Saturdays to reach the top layer, there is a session of its own.

One sentence, if you remember nothing else

AI makes you faster at whatever you already are. If you are a careful marketer who understands the business behind the page, it will multiply that. If you are guessing, it will help you guess at ten times the speed, in better English, with more confidence, for a client who is already drafting the message that ends the contract.

Learn the work. Then let the machine carry what it is good for.

Kahill Digital Marketing School trains digital marketers online across Uganda and beyond, and works with the businesses hiring them. school.kahillinsights.com