Learning to use AI when you're starting from zero

You've been hearing about artificial intelligence for two years, everyone around you seems to use it, and you still haven't taken the plunge. It isn't a question of ability: it's a question of vocabulary. Most content assumes you already know what a prompt, a model or a token is, and you lose the thread by the third sentence.

This article starts at square one. No technical knowledge, no paid account and, above all, no promise of a life transformation. Just what you need to open a tool and get something useful out of it today.

What generative AI is, without metaphors

A generative AI like ChatGPT, Claude or Gemini is a program that has read an enormous amount of text and has learned, through sheer repetition, to predict which word comes after which. When you type a question, it produces an answer word by word, each time picking the most plausible continuation given everything that came before.

That description sounds reductive, yet everything follows from it. It explains why these tools write so well, why they're excellent at rephrasing, summarising and translating — and why they can state something false with exactly the same confidence as something true. The program doesn't know it's wrong: it produces what looks like a good answer.

The ten words that trip everyone up

You only need the first four to get started. The others will let you read an article without getting lost.

Your first conversation, step by step

Pick any mainstream tool; the free versions are more than enough to explore. Then do this exercise, which takes five minutes and teaches you more than three articles.

1. Ask a question you already know the answer to. Something about your job, your town or your daily life. You'll see right away how accurate it is — and, above all, that it sounds just as sure of itself when it's wrong.
2. Ask for a rewrite. Paste something you wrote and ask it to make it clearer without changing the meaning. This is where these tools shine brightest.
3. Contradict it. Tell it that it made a mistake, even if it didn't. Watch what happens: it often apologises and changes its answer. That's the most important lesson about its reliability.
4. Ask it to keep it short. By default it writes long, structured answers. “In three sentences” changes everything.

By the end of these four steps, you'll know more about its strengths and limits than most people who use it every day.

Five uses that help from day one

Nothing spectacular: five tedious tasks that take a tenth of the time.

• Understanding an official letter. Paste it in and ask what's expected of you and by when.
• Writing a difficult message. A follow-up, a polite refusal, a complaint. Give the facts, ask for the tone.
• Summarising a long text to decide whether it's worth reading in full.
• Drafting a list — groceries for a week of meals, gear for an activity, questions for an appointment.
• Explaining a concept at your level. “Explain this as if I knew nothing about it” works remarkably well.

Notice what they have in common: in all five cases, you can still judge the result yourself. That's the rule that protects you from everything else.

The five beginner mistakes

1. Asking something too vague. “Tell me about marketing” gets you a generic lecture. “I sell handmade candles at a local market — what are the three cheapest ways to get known in my area?” gets you something you can use.
2. Giving no context. The AI knows nothing about you. Who you are, who it's for and what it's for: those three details change the answer completely.
3. Starting over every time. Instead of rewriting, say “shorter”, “more direct”, “drop the third part”. The conversation builds on itself.
4. Trusting numbers and sources. That's precisely where it invents the most. Any date, statistic or reference needs to be checked elsewhere.
5. Hunting for the magic prompt. Lists of “100 life-changing prompts” are almost useless: what matters is how you phrase things, not the formula.

That last point deserves an article of its own — it's the subject of our guide to writing good prompts.

When to trust it, when to check

One simple rule sums up the table: use it to transform information you provide, and be wary when you ask it to provide the information.

Where to go next

Once the fear is gone, there are two things left to learn, in this order: understanding why it gets things wrong, then learning how to phrase requests. That's exactly the progression of the Nexily AI Academy, whose Level 0 starts right where you are today.

Don't skip ahead to automation: people who start with advanced tools almost always have to go back, because they're missing the basic mental model.

Frequently asked questions