How to start with AI: begin with something you actually enjoy
The best way to start with AI is not a list of courses or a perfect prompt. Begin with a small project you enjoy. That is where you will learn fastest.
Visual guide
ChatGPT + Claude
Imagine your first attempt
Your first encounter with AI does not have to be theoretical. It could look like an architect standing in a studio, with a luxury Beverly Hills villa floating in the space ahead, changing its materials, light, terrace and view by hand.
Not because you want to become an architect. Because a scene like this immediately teaches you how AI works: set a direction, review the result, say what should change and try another version.
Try this first prompt: Design a luxury cliffside villa in Beverly Hills. Modern style, plenty of glass, warm stone, a city view and evening sun. First, give me three visual directions. Then turn the strongest direction into a detailed image prompt.
Play with the next version
The first result does not have to be great. In fact, it is better when it is not. It shows you what was missing from the brief.
Perhaps the villa looks too much like a catalogue image, so you add a more cinematic atmosphere. Perhaps the scene feels empty, so you add material samples, brick, stone, glass, wood and lights. Perhaps the result is too science-fiction, so you write: make it more realistic, like a photograph from an architecture studio.
Try
Notice
Improve
That is the whole trick. You do not learn AI by reading rules. You learn by experiencing them in a small attempt that matters to you.
Choose your own direction
It does not have to be architecture. What matters is choosing a subject that pulls you in. If you enjoy marketing, create an ad. If you enjoy video, invent a scene. If you enjoy business, outline a small product. If you enjoy writing, let AI develop an idea and then rewrite it in your own voice.
Marketing
Video
Business
Writing
Open one tool
You do not need ten apps at the beginning. ChatGPT, Claude or Gemini is enough for text and thinking. For images, use whichever visual tool you have at hand. Video and automation can come later.
The key is not knowing every tool. The key is starting to create.
A small task for today
Choose one idea and give AI the first brief. Do not wait for a perfect prompt. Make the first sketch, find one thing you want to change and ask for another version.
After a week, you may not know every model. But you will have your own experience. That is more valuable at the start than ten saved guides you never return to.
Sources
Frequently asked questions
What people often ask
How should I start with AI if I know almost nothing about it?
Start with one small project that genuinely interests you. It could be a poster, a piece of writing, a simple travel plan, a room design, an advertising idea or a personal assistant for your notes. You do not need to complete an entire course first or know the difference between every model. Choose one tool, give it a concrete task, review the result and write the next prompt according to what you want to change. That is how you begin learning naturally.
Do I need to learn how to write perfect prompts first?
No. Good prompts develop gradually. Your first brief can be ordinary and imperfect. What matters is reviewing the result and naming what should change: make it more specific, shorter, bolder, simpler or closer to your voice. The more you iterate, the more you understand how AI responds to context, examples and constraints. Prompting is not a magic formula. It is a conversation about the result.
Which AI tool should I choose first?
For text, ideas and everyday work, start with ChatGPT, Claude or Gemini. For images, try Midjourney, ChatGPT image generation or another visual tool you already have access to. Do not worry about the perfect stack yet. Your first tool only needs to be good enough to get you into practice. Once you hit a limitation, you will understand what you need next: better image work, longer context, automation or a connection to data.
What if the AI output disappoints me?
That is normal and part of the process. The first output often mainly shows what was missing from the brief. Treat it as a sketch, not a final answer. Ask what is usable, what is off and what context you still need to provide. Then adjust the prompt, add an example, change the style or split the task into smaller steps. This loop of attempt, feedback and another attempt is where most AI learning happens.
Keep going
Related articles
More guides from the same area, topics and tools.
How to build an AI copywriting workflow that saves hours without damaging the brand
AI copywriting in 2026 depends on an editorial system with brand voice, claim guidelines, a curatorial filter and a human signature. Here is how to build it step by step.
What are AI skills, and why are they the future of work with AI agents?
Skills turn AI tools such as Claude Code, Codex and Cursor into repeatable systems that apply your know-how consistently. Learn the principle, see three strong GitHub repositories and inspect a real editorial skill that helped produce this article.
How to build an AI creative system that keeps the brand consistent across every format
Modern banner production is not a workflow inside one tool. It is a system: a master layout, brand rules, format rules and an AI skill that applies them repeatedly. Here is how to produce eight formats every week without chaos.