Skip to main content
Learning roadmap

Three stages to turn AI into a real work skill

There's no shortage of AI tutorials online — the problem was never the resources, it's the order. This roadmap sets the order for you: what to learn at each stage, what counts as done, and where to go next.

1

Stage 1: Understand it, dare to ask

Weeks 1–2 · 15 minutes a day

Goal of this stage: Build an accurate mental model: know what AI can and cannot do, and feel comfortable asking.

What you'll learn

  • What large language models actually do (with analogies, no math)
  • Why AI states wrong things so confidently: where hallucinations come from
  • What tokens, context and the conversation window really mean
  • The four-part prompt: role × task × context × format
  • Your data safety line: what should never be pasted into AI

What counts as done

  • You can explain "why AI gets things wrong" to a colleague in your own words
  • You can turn a vague request into a clear task description
  • You instinctively verify any figure or legal clause AI gives you

Getting started: 20 minutes to your first win · AI basics · Safety, privacy and judgment

2

Stage 2: Use it fluently, save real time

Weeks 3–8 · 15–30 minutes a day

Goal of this stage: Put AI into the work you actually do every day, so the time you save is visible and easy to explain.

What you'll learn

  • Build 5–10 prompt templates tailored to your own job
  • Pick the right tool: writing, slides, images, video, research
  • Handle files and long documents: summarise, compare, pull out what matters
  • Organise information into tables, outlines and action lists
  • Judgment: which tasks to hand to AI and which you must do yourself

What counts as done

  • You have your own prompt notes you can copy straight into similar work
  • At least two weekly routines now take less than half the time
  • You can clearly explain "why I don't hand this task to AI"

Prompt library · AI tool guide · Work and life use cases

3

Stage 3: Build workflows, scale your output

Week 9 onwards · project by project

Goal of this stage: Go from asking one question at a time to a whole workflow that runs itself, and bring your team along.

What you'll learn

  • Build a personal or team knowledge base so AI answers from your own material (the idea behind RAG)
  • Get to know AI agents: when to let AI run several steps on its own
  • Connect automation tools: turn repetitive work into scheduled jobs
  • Build small tools without coding: an intro to vibe coding
  • Roll it out to your team: usage guidelines, shared prompts, training

What counts as done

  • You have a workflow that runs every week without you re-prompting it
  • You've built a small tool or automation script you actually use
  • You can design AI usage guidelines and a prompt library for your team

Workflows and automation · Intro to vibe coding · Glossary

Four principles for the journey

Whatever stage you're at, these four apply.

Start with real work

Don't learn AI for its own sake. Pick something you already have to do this week and try it with AI — that's what makes the learning stick.

Change one variable at a time

Switch tools, models or phrasing one at a time. Otherwise you'll never know which change made the result better.

Save the prompts that work

When you get a great answer, save that prompt to your notes right away. Three months later those notes beat any tutorial.

Always keep the final call

AI handles drafts and speed; you handle facts and responsibility. Your name is on it, so verifying is your job too.

Not sure which stage you're in?

If you haven't yet finished one real piece of work with AI, start at stage 1. If you already use it every day but don't feel any faster, what you need is stage 2: prompts and picking the right tool.