To learn AI from scratch, start with AI literacy and prompting, then build simple projects with no-code tools, learn Python, build AI apps using model APIs, and finally study machine learning foundations if you want to train models. Most beginners can build useful AI projects within three months with steady practice.
First, decide what "learning AI" means for you
"Learning AI" can mean very different things:
- Using AI well in your studies or job.
- Building with AI: automations, chatbots and apps that use existing models.
- Creating AI: training and researching models, which requires serious maths and programming.
Most people need the first two. Only some will want the third. The roadmap below takes you through all three in order, and you can stop at the stage that fits your goals.
Stage 1: AI literacy (weeks 1 to 2)
Understand what AI is, how large language models work, why they hallucinate and how to use them responsibly.
- Read: what is AI, how LLMs work.
- Practise: use an AI assistant daily for real tasks and check its answers.
Stage 2: prompting (weeks 2 to 4)
Learn to write clear, structured prompts and evaluate output.
- Read: prompt engineering for beginners.
- Practise: build a library of prompts for your studies or work.
Stage 3: build without code (weeks 4 to 8)
Make real things using no-code tools and AI features:
- a custom assistant trained on your notes,
- an automation that connects a form, a spreadsheet and email with an AI step,
- a simple website or app built with an AI builder.
This stage teaches how AI behaves inside real systems. See our AI automation and no-code courses.
Stage 4: Python (months 2 to 4)
Python is the main language of AI. Learn variables, loops, functions, lists, dictionaries, files and calling APIs. Build small programs that call an AI model.
- See Python projects for beginners and our Python tutoring.
Stage 5: build AI applications (months 4 to 6)
Use model APIs to build applications: a chatbot with memory, a document Q&A tool using RAG, a simple agent. Learn testing and deployment. This is the core of our generative AI development course.
Stage 6: machine learning foundations (months 6 onwards)
If you want to understand and train models: statistics, linear algebra basics, pandas, scikit-learn, model evaluation, then neural networks. See machine learning and data science.
Stage 7: specialise
Choose a direction: computer vision, natural language processing, data science, AI engineering, AI product management, or AI in your own field such as healthcare, finance or education.
Typical timeline
| Goal | Stages | Realistic time with steady practice |
|---|---|---|
| Use AI well | 1 to 2 | 1 month |
| Build automations and simple tools | 1 to 3 | 2 to 3 months |
| Build AI apps with code | 1 to 5 | 6 to 9 months |
| Machine learning foundations | 1 to 6 | 12 months or more |
Tips that make the difference
- Build from week one. Projects teach faster than courses alone.
- Learn in public. Share projects on GitHub or LinkedIn.
- Get feedback. A mentor saves months of confusion.
- Avoid tutorial hell. Watching is not learning; building is.
Frequently asked questions
Can I learn AI without a computer science degree?
Yes, for using and building with AI. Research roles typically require advanced degrees.
How much maths do I need?
None for stages 1 to 3, little for 4 and 5, and solid statistics and algebra for machine learning.
Is it too late to start?
No. The tools are new for everyone, and beginners who practise consistently progress quickly.
