AI vs Machine Learning vs Data Science: The Difference

McTaba runs coding and data bootcamps in Kenya. This guide is written to be useful whether or not you ever join one.

Open your phone. A text arrives: "Possible fraud on your M-Pesa. Reply STOP." A lending app offers you a limit within seconds. A chatbot on a bank's WhatsApp number answers in plain English. People call all three "AI". Engineers would split them into different layers.

The easiest picture is three circles, one inside another. The big circle is artificial intelligence. Inside it sits machine learning. Data science overlaps both, but it is a job and a method more than a technology.

If you are deciding on a career, the wider map is in the data science career guide for Kenya. This page only sorts out the words.

Term	Plain meaning	Kenyan example
Artificial intelligence (AI)	Software that does something we would call smart if a person did it	A WhatsApp chatbot that answers a customer's question about their loan balance
Machine learning (ML)	A way of building AI by letting a program learn patterns from past examples	An M-Pesa fraud alert that learned from past scams which transactions look odd
Data science	The work of collecting, cleaning and studying data to answer a question or build a model	An analyst who studies a chama's records to find which members repay late

Notice the overlaps. The fraud alert is AI (it acts smart) and ML (it learned). The data scientist who built it spent most of the time on data science: collecting examples, cleaning them, checking the results.

Not all AI is machine learning. An old rule-based chatbot that matches keywords is AI of a sort, yet it learns nothing. And not all data science is machine learning. Much of it is counting, charting and explaining, with no model at all.

For a worked example of ML on a small problem, see machine learning explained with a matatu route.

If you want to learn all three layers in order, our Data Science programme in Nairobi is built that way.

Take a mobile lending app. It looks at a borrower's phone history and decides a limit in seconds.

• Data science is the earlier work: gathering past borrowers' records, cleaning errors, asking which signals relate to repayment, and deciding how success is measured.
• Machine learning is the model trained on that data. Given a new borrower's signals, it outputs a repayment score.
• AI is the whole product behaving smartly: the app that reads the score, sets a limit and talks to the customer.

The same layers show up in chama credit scoring, where the records are a group's savings rather than a phone's history.

Today the word AI is also used for large language models such as the ones behind chat assistants. Those are a kind of machine learning trained on text. Building products on top of them is often called AI engineering. Our programme covers it in the Intro to AI Engineering phase.

Job titles blur. Employers mix them, so read the duties in each advert, not the title. A rough split:

• Data analyst. Answers questions with SQL, spreadsheets and dashboards. Less modelling, more explaining.
• Data scientist. Cleans data, runs experiments and builds models, then explains what they found.
• Machine learning engineer. Takes a model and makes it run reliably behind an API, with monitoring. More software, less research.
• AI engineer. Builds products that call language models, retrieval systems and agents. Heavy on software and testing.

For how analyst, scientist and engineer roles compare in Kenya, read data analyst vs data scientist vs data engineer. For pay, use dated salary pages rather than guesses. No school can promise a job, and titles that sound grand do not guarantee better work.

Start with data science basics, because every other layer rests on them. In order:

1. Spreadsheets and SQL. Ask questions of a table.
2. Python and pandas. Clean and reshape data.
3. Basic statistics. Know when a pattern is real or luck.
4. Machine learning. Train, test and measure a model.
5. Deployment and AI engineering. Put a model behind an API or a WhatsApp reply.

Skipping to step 5 is tempting, since AI headlines are loud. People who do it tend to build demos that break on real, messy Kenyan data. A route that avoids that is laid out in the data science roadmap for beginners.

Our own programme follows the same order across 9 phases, from Python to a deployed capstone, over 5 months full time or 8 months part time. The next intake is 11 January 2027, and the programme asks for no coding, data or maths background needed. See the Data Science programme in Nairobi for fees and the syllabus.