What is the best way to learn AI from scratch?

If you want to use AI and not build it, start with a chat tool and a small task from your own week. Then learn to ask clearly, check the answers against a source you trust, keep the prompts that work and finish one small real thing. Practising in that order teaches you more than reading about AI.

Do I need to learn to code or know maths to learn AI?

Not to use it. Maths and coding matter if you want to build AI models. If you want help with emails, plans, writing or a small project, plain English is enough.

Can I learn AI for free?

Yes. Free AI chat plans exist, and every lesson on this site can be read without an account. An optional account saves your progress and gives you a certificate of completion, which is not a qualification. A vendor's free plan can change, so check its own pricing page before you rely on it.

How long does it take to learn the basics of using AI?

There is no fixed answer. It depends on you and what you want to use AI for. Take it one lesson and one real task at a time, and use the checkpoints to tell when you are ready to move on.

Is a short course better than a long one for a beginner?

For most beginners, a short course you can finish is the better choice. You get more from a course you complete and use than from a long one you give up on. Each lesson here takes 5 to 15 minutes and every course ends with a finished thing.

Best Ways to Learn AI If You Do Not Code

Short answer: if you have no plans to write software, you can skip the maths and the coding. Learn in this order: use a chat tool, ask it clearly, check what it says, keep the prompts that work, then finish one small real thing. Each step below ends with a checkpoint, so you know when to move on.

First decide: do you want to use AI or build it?

Most advice skips this question, and your answer changes everything that follows.

Using AI means you open a chat tool such as ChatGPT or Claude, type what you need and get help with an email, a plan, a pile of notes or a first draft. Building AI means making the tools themselves: training models, writing the code around them and testing how well they work. That is a different job, and it needs different skills.

So before you open a tutorial, write down what you want AI for. Replying to customers faster? Planning the week’s meals? A simple web page for a club? Those are all “use” goals, and this post is written for them.

Why the popular roadmaps start with maths and Python

If you search for how to learn AI, most of the top results are roadmaps from course providers, universities and video channels. They usually start with maths, move on to Python and then to machine learning.

Those roadmaps do their job. If you want to build models, you need those foundations, in the same way a mechanic needs to understand engines. If you only want to drive, starting with the engine is a long detour, and you may give up before you use the tool you came for.

Our position is simple. For people who do not code, practising on a real task beats reading about AI.

Step 1: Get comfortable talking to an AI chat tool

Pick one general AI chat tool and give it something small from your real week. For example:

Draft a reply to an awkward message.
Turn a rambling note into three clear points.
Explain a term from a document you have been putting off.

Don’t worry about a perfect answer yet. Get used to the back and forth. You can reply “shorter”, “friendlier” or “no, I meant the other thing” in the same chat, and it changes the answer. Before you type anything personal, know what to keep out: bank or card details, health records, passwords and confidential documents.

Checkpoint: you have used it on one real task and improved the answer by replying in the same chat, without starting again.

AI for beginners: start here covers this step: what these tools are and what they get wrong, how to make a free account and choose your privacy settings, and your first conversation.

Step 2: Learn to write a better prompt

Once you are comfortable, you will notice that the answers vary a lot. Mostly that comes down to how you asked. A vague request gets a vague answer, because the AI fills the gaps with guesses. OpenAI’s own advice is to be clear and specific, give enough context, then review the reply and refine your prompt.

A useful shape has four parts:

Task: what you want it to do.
Context: who it is for and why.
Format: how long it should be and how it should be laid out.
Rules: the tone, and anything to leave out.

If you can brief a colleague well, you already have most of this skill.

Checkpoint: you can take a prompt that gave you a so-so answer, rewrite it with all four parts and get a clearly better one.

Prompting for beginners works through this with the exact prompts we used and the replies that came back, so you can compare your attempts with real examples.

Step 3: Learn to check what it tells you

People tend to rush this step. AI chat tools can sound sure of themselves and still be wrong. Anthropic and OpenAI both say so on their help pages: answers can include wrong facts and dates, made-up quotes or references, and confident replies that are mistaken. OpenAI tells users to check important information against reliable sources. Anthropic says not to rely on Claude as your only source of truth.

How to check:

Pick out the claims that would matter if they were wrong: names, dates, figures, quotes and prices.
Go to whoever publishes the real thing, such as GOV.UK for bank holidays or the shop itself for opening hours. Asking another AI doesn’t count. That is just marking its own homework.
Check every line, and keep going after the first few.

Treat each answer as a first draft from a keen assistant who sometimes makes things up.

Checkpoint: you have compared one answer, line by line, with an official source.

The lesson Check what it tells you walks through this with a real question and the official page it was checked against.

Step 4: Keep the prompts that work

This is where a lot of time goes missing. You write a prompt that works, use it once, then write something similar from scratch the following week.

Keep one document instead. When a prompt works, paste it in with a line saying what it is for. Then replace the parts that change each time with square brackets, like [paste the email here]. Keep the parts that stay the same, especially your rules, because they do most of the work. Next time, you only fill in the brackets.

Checkpoint: you have reused a saved prompt on a second real job.

Build your prompt kit is the lesson that turns your three most useful prompts into templates, each tested on a real job.

Step 5: Finish one small real thing

Reading about AI can feel like progress. Finishing something is real progress.

Pick one small thing you would actually use: a week of evening meals with a shopping list, or a one-page website for a club or small business. The last lesson of AI for beginners ends with a meal plan you have checked yourself. Build a website with AI takes you from a short brief to a page that is live online, and there is no code to learn: you copy, paste, check and drag a folder.

Checkpoint: it is finished, you have checked it, and someone else could use it or look at it. Done matters more than perfect.

What to skip for now

Maths, until you want to build models.
Python as a first step, unless you want to learn it for its own sake.
Huge “everything about AI” playlists. They feel productive and rarely end in anything you made.
Endless tool comparisons. Choosing between tools can use up an evening you could have spent using one.

And skip the worry that you are behind. One lesson at a time is a good pace.

How to tell a course is worth your time

It is short enough to finish, and each lesson fits in one sitting.
It shows the exact prompts and the real replies, along with when it got them.
It ends with a finished thing you can use, and you can say exactly what you covered.
It says what the tools cost and links to where that came from.
It makes no promises about jobs or results.

Be clear about what a certificate means, too. A certificate of completion shows you finished a course. It is not a qualification.

Our courses follow this pattern. There are three live courses, every lesson takes 5 to 15 minutes, each course ends with a finished thing, and every lesson can be read without an account.

The best ways to learn AI if you do not code