From beginner to pro

Your level isn't measured by what you know about AI. It's measured by what's left after the prompt: rework, or a finished result. Plenty of people have sat in level one for a year and assume they're at level three — they just never checked.

There are three levels: use, prompt, and automation. Each has a typical task, a sign you've moved past it, and a next step. Order matters: skipping a rung is what causes most of the frustration.

Level 1: everyday use

Here the request is short and the tool is a tab you keep open all day: summarize, rewrite, explain, list, unblock a stuck email. Most people stay at this level — and it's a fine place to start.

The task is to pick 3 kinds of work you repeat every week and start doing them with AI: meeting notes, a message to the team, a rough draft. See 5 prompts that save hours if you want starting points.

Sign you've moved up: you open AI for every repetitive task without weighing whether it's worth it. It became reflex.

Level 2: structured prompts

At level 2 the request stops being a conversation and becomes a specification: who the AI is, what the context is, what format should come out, and what must not appear. That's the difference between "write an email" and a prompt that arrives ready to use.

Required reading: OpenAI's prompt engineering guide. And when the task is large, break it up with the Ladder Rule: asking for everything at once is what produces generic answers.

Sign you've moved up: the first response already arrives in the right format, and you stop correcting the same thing twice.

Level 3: automation and flow

At level 3 the prompt isn't the destination — it's one step. You build a fixed context document (who you are, your product, your market), paste it at the start of every new conversation, and chain the rest: raw data, analysis, draft, review.

It's worth understanding what's underneath: what machine learning is explains why the model fails and where it's reliable. The next rung is multi-agent workflows — tasks split across steps without you walking the same path twice.

Sign you've moved up: you no longer rebuild the same request every week — and when the output affects a person, a human check happens before anything is published or decided.

Signals that you've moved up a level

• Time: the same task takes minutes, not rounds of correction.
• Corrections: you edit the content, not the format.
• Reuse: your prompts became templates other people use.
• Limits: you know off the top of your head when not to use AI.

If two of those four don't hold, drop back a level and repeat for a week. Progress with AI is boring on purpose: repetition with context, not a shortcut.

Frequently asked questions

Do I need to know how to code to reach level 3?

No. Level 3 is about sequence and context, not code. Plenty of tools chain steps in a visual flow without a single line of code — what changes is how the process is designed.

Which tool should I pick?

The one you open every day. Switching tools won't raise your level: ChatGPT, Claude, and Gemini all cover levels 1 and 2 well. Pick one, standardize your prompts, and only reconsider if the task demands it.

How much time a week does this take?

One new task per week already moves the needle: one task that's manual today becomes a template, and a template becomes a step in a flow. The bottleneck is rarely time — it's never deciding which task to automate first.

Does automating take away my role?

It removes repetition, not judgment. At level 3, the work that's left is the best part: deciding what to verify, reading the output, and standing behind the decision — something a model doesn't do.

For a direct continuation, go to 10 ways to use AI at work — the same track, with examples ready for the week.