“Learn AI” is a strange thing to tell a beginner, because AI isn’t one skill — it’s a stack of very different things: using AI tools well, understanding how the models work conceptually, and (for a smaller group) actually building with AI programmatically. Most beginner guides blur these together and end up vague. This roadmap keeps them separate so you know exactly what you’re learning and why, month by month.
First, decide which track you actually need
Before picking a course, be honest about your goal:
- “I want to use AI tools better at my current job” — you need Track 1 only (below), and you can be genuinely competent in 3-4 weeks of consistent use.
- “I want to add AI skills to make myself more hireable / freelance” — you need Track 1 plus Track 2, roughly 2-3 months.
- “I want to build AI-powered products or work as an AI/ML engineer” — that’s a genuinely different, much longer path requiring programming fundamentals (Python), and it typically takes 6-12+ months of real study, not a side-hustle-style roadmap. Be wary of anyone promising otherwise.
Track 1: Using AI Tools Competently (Weeks 1-4)
Week 1 — Get hands-on with the core chat tools. Use ChatGPT, Claude, and Gemini for real tasks you already do — drafting emails, summarizing documents, brainstorming. Don’t just read about prompting; do it daily on real work so it becomes a habit rather than a novelty.
Week 2 — Learn actual prompting technique, not tricks. The skill that matters is being specific: giving the model context (who you are, what you want, constraints), asking for a particular format, and iterating on its output rather than accepting the first draft. Anthropic and OpenAI both publish free prompting guides worth reading once you have hands-on experience to anchor them to.
Week 3 — Learn the limits. Deliberately test where these tools get things wrong: ask about very recent events, obscure facts, or math that requires careful multi-step reasoning, and see where the answers go wrong or need a tool with live search/browsing. Knowing where AI is unreliable is as valuable as knowing where it’s useful.
Week 4 — Pick 2-3 tools relevant to your actual work (writing, spreadsheets, design, coding, research) and go deep on those specifically rather than trying to sample everything.
Track 2: Becoming Genuinely Useful With AI at Work (Months 2-3)
Month 2 — Build real workflows, not one-off prompts. The professionals who stand out aren’t the ones who know the most tools — they’re the ones who’ve built a repeatable process: a specific sequence of prompts and tools for a task they do often (writing reports, analyzing data, preparing client materials). Document your own workflow once you find one that works.
Month 3 — Learn to evaluate AI output critically. This is the skill almost nobody teaches: knowing when an AI-generated answer is confidently wrong, checking claims against a real source, and editing AI drafts so they don’t read as generic. This is also, not coincidentally, the exact skill that separates content that gets flagged as low-value from content that reads as genuinely useful.
What I’d skip
Be skeptical of “become an AI expert in 30 days” courses and of vague promises about salary jumps — actual outcomes depend heavily on your existing role, industry, and how you apply the skill, not the AI knowledge alone. Free resources (official documentation from Anthropic, OpenAI, and Google; YouTube channels that show real work rather than hype) will get most beginners further than a paid course, at least for Track 1 and Track 2.
If you want to go further: the technical track
If your goal is actually building AI systems — not just using them — you’ll need programming fundamentals first (Python is the standard starting point), then a structured course in machine learning basics before touching generative AI specifically. This is a real commitment measured in months, not weeks, and skipping the fundamentals to jump straight to “AI” tends to produce people who can talk about AI but can’t build anything reliable.