Getting Into AI: What You Actually Need to Know

AI is the most talked-about technology on the planet right now. Every company is rushing to integrate it, every university is adding AI courses and every LinkedIn post promises you can "learn AI in 30 days." The noise is deafening and for someone trying to figure out where to actually start, it's paralysing.

This article cuts through the hype. Here's what AI actually involves, what skills you genuinely need and the realistic path to making it your career.

First: What "AI" Actually Means in a Job Context

When people say "AI jobs," they usually mean one of several different roles and they require very different skill sets:

• Data Analyst, Analyses existing data to find patterns and produce insights. Uses SQL, Excel, Python (Pandas, Matplotlib). Less math-heavy, more accessible as a starting point.
• Machine Learning Engineer, Builds and trains models that make predictions or decisions. Requires strong Python, statistics and ML libraries (scikit-learn, TensorFlow, PyTorch).
• AI Engineer / LLM Engineer, Integrates large language models (like GPT) into applications. Hot right now. Requires Python, API knowledge and an understanding of prompt engineering.
• Data Scientist, A blend of analyst and ML engineer. Explores data, builds models and communicates findings to business stakeholders.

The good news: most of these roles share a common foundation. Get the foundation right and the specialisation becomes much easier.

The Core Foundation (What Everyone Needs)

1. Python

Python is the language of AI and data science. It's readable, powerful and has the best library ecosystem for machine learning and data analysis. If you don't know Python, start here. You don't need to be a Python expert before moving to AI, but you do need to be comfortable with variables, loops, functions, lists, dictionaries and working with files.

2. Statistics & Probability Basics

You don't need a maths degree. But you do need to understand concepts like mean, median, standard deviation, probability distributions and correlation. These come up constantly when you're evaluating whether a model is actually working. Most of this can be learned in 2–3 weeks with a focused resource.

3. Pandas & NumPy

These are Python libraries for working with data. Pandas gives you DataFrames, essentially supercharged spreadsheets in code. NumPy handles numerical operations efficiently. You'll use these in almost every data or ML project.

4. Data Visualisation

Being able to visualise data is essential, both for understanding it yourself and for presenting findings to others. Matplotlib and Seaborn are the standard tools. Plotly is great for interactive charts.

Machine Learning: The Layer on Top

Once you have the foundation, machine learning starts to make sense. ML is essentially teaching a computer to make decisions by showing it lots of examples. The main concepts to learn:

• Supervised learning, training a model on labelled data (e.g., spam vs. not spam). This is where most beginners start.
• Unsupervised learning, finding patterns in data without labels (e.g., clustering customers by behaviour).
• Model evaluation, understanding accuracy, precision, recall and how to know if your model is actually good.
• scikit-learn, the go-to Python library for traditional ML. Clean API, excellent documentation and covers the majority of real-world ML use cases.

The LLM / AI Engineering Wave

Large language models (LLMs) like GPT-4, Claude and Gemini have created an entirely new category of AI engineering. Companies want developers who can integrate these models into products, chatbots, document summarisers, code assistants, automated workflows.

This is genuinely more accessible than traditional ML. You don't need deep statistics or model training knowledge. You need:

• Python and API knowledge
• Understanding of how to write good prompts (prompt engineering)
• Familiarity with tools like LangChain, OpenAI SDK, or Anthropic SDK
• Knowing how to evaluate and improve model outputs

If you're coming from a web development background, this is the fastest way to enter the AI space.

What the SA Market Is Hiring For

Looking at data and AI job listings in South Africa, the most common requirements for entry-level and junior roles are:

• Python (non-negotiable)
• SQL and experience with databases
• Pandas, NumPy and basic data visualisation
• Understanding of ML concepts (even without deep expertise)
• A portfolio project that demonstrates real analysis or model building
• Communication skills, can you explain your analysis to a non-technical person?

In financial services (Discovery, Investec, Standard Bank), insurance and retail, all major SA employers, data analysts and data scientists are among the most in-demand hires. These industries sit on enormous amounts of data and are all aggressively building AI capability.

The Realistic Timeline

• 3–4 months: Comfortable with Python, Pandas, visualisation and basic ML. Ready to apply for data analyst roles.
• 6–9 months: Solid ML foundations, experience deploying a model, strong portfolio. Ready for ML engineer or data scientist applications.
• Ongoing: AI moves fast. The best AI practitioners treat learning as permanent, not a phase you complete.

The Common Mistakes to Avoid

• Jumping straight to deep learning, understand the basics first. You can't debug a neural network if you don't understand logistic regression.
• Tutorial paralysis, watching courses without building anything. Every week you should produce something, even if it's small.
• Ignoring business context, the best data scientists understand the problem they're solving, not just the technique they're using.
• Skipping the fundamentals, Python and statistics done properly will save you months of confusion later.

Your Best First Step

If you're starting from scratch: begin with Python. Spend 4–6 weeks getting comfortable with the language before touching any AI or data libraries. Then move to Pandas and basic data analysis. Then machine learning.

If you already code in another language: your transition to Python will be faster. Focus on the data science-specific libraries and start building projects with real datasets immediately.

If you want a guided path with mentorship at every stage, our AI & Data Science track at The Developer Culture is built exactly for this journey, starting from fundamentals and taking you all the way to deployment.