Most "how to learn AI" advice is useless. Take a course. Try ChatGPT. Experiment. Thanks, very helpful.
What nobody tells you is that AI fluency isn't a destination — it's a progression. And most people get stuck at level one thinking they've arrived. They've visited. They haven't moved in.
Over the Super Bowl weekend earlier this year I went from 2 agents to 11 running continuously on my home server in 72 hours. Financial monitoring, family logistics, flight parsing, automatic debugging — all of it. I'm not telling you that to brag. I'm telling you because 18 months ago I was asking ChatGPT to clean up my emails. The distance between those two things isn't talent. It's deliberate progression.
Here's the roadmap I wish someone had handed me.
Level 1: Foundation — Stop Being Afraid of It
Time investment: 2–4 weeks, 30 minutes a day
Most people at this stage are either avoiding AI entirely or using it so tentatively that they're not getting anything useful out of it. The goal here isn't mastery. It's removing the friction and building the habit.
Start with one tool. Just one. Claude, ChatGPT, Gemini — pick one and use it every single day on real work problems. Not toy problems. Not practice prompts from a course. Your actual work.
What to focus on:
- Ask it to draft things you'd otherwise write from scratch
- Ask it to summarize documents you don't have time to read fully
- Ask it to explain things you don't fully understand
- Ask it questions you'd normally Google
The most important thing at this stage is learning that the quality of your output is directly tied to the quality of your input. Vague questions get vague answers. Specific, contextualized questions get useful ones. Start paying attention to that cause and effect.
Level 2: Practitioner — Make It Part of How You Actually Work
Time investment: 1–3 months of intentional practice
This is where most people plateau and don't realize it. They're using AI regularly but for the same three things. Emails, summaries, brainstorming. The tool is in their workflow but their workflow hasn't fundamentally changed.
The shift at this level is moving from "AI helps me do my tasks" to "I'm rethinking which tasks I should be doing at all."
What to focus on:
- Prompt engineering — learn to give context, set constraints, define the output format you want.
- Chaining — use the output of one AI interaction as the input for the next.
- Domain-specific application — build prompts tailored to your actual job.
- Discernment — check outputs the way you'd check the work of a junior colleague before presenting it.
The people I see at this level who are really accelerating are treating their best prompts like assets. They're saving them, refining them, sharing them with their team.
Level 3: Transformational — Rebuild What You're Capable Of
Ongoing — this is a practice, not a finish line
At this level you're not using AI to do your job better. You're using AI to do things that weren't possible before. You're an orchestrator — you direct, judge, and build systems — not just someone who completes tasks.
What this looks like in practice:
- Building workflows, not just using tools
- Agentic thinking
- Leverage
- Building AI literacy in your team
What separates someone at this level isn't access to better tools. It's intellectual courage — the willingness to genuinely challenge how they work, not just augment it.
The honest thing most roadmaps won't tell you
There's no shortcut through Level 1 and 2. I see people try to jump straight to "transformational" and they fall apart because they haven't built the foundational judgment to know when the AI is wrong, when to push back, and where the guardrails need to be.
The calculator analogy works here. An accountant who understands numbers uses a calculator to go faster. An accountant who doesn't understand numbers uses a calculator to make mistakes with more confidence. The tool amplifies your judgment — which means you need to build the judgment first.
And the other thing nobody says: this takes time. Thirty minutes a day for six months will get you further than a two-day AI bootcamp and then nothing. Consistency beats intensity.