The AI Learning Path: What to Focus on First, Second, and Third

Most people try to start learning AI the same way they’d approach a new app: find the feature list, click through the menus, figure out what everything does. That approach works great for Excel. It doesn’t work for AI. There’s so much it can do that lies beneath the surface, and it’s activated through the completely free-form “menu” of a conversation.

This is the path we walk people through in our Getting Started series. Not the only path, but the one that consistently gets mid-career professionals from “I’ve barely touched this thing” to “I use this every day and it’s changed how I work.” Three phases, built on each other.

Phase 1: Get a Repeatable Win

The single goal of phase one is not to “understand AI.” It is to have one experience where AI saves you real time on real work. One. That’s it.

This matters because motivation runs on evidence, not theory. Reading about AI is not the same as using AI. The best predictor of whether someone builds a lasting habit with these tools is whether they got a concrete win in the first two weeks.

What to do in phase one

Pick one task you do regularly, that has a clear output, and that you’ve done enough times that you’d know immediately if the result was good or bad. Good candidates: drafting a routine email, summarizing a meeting agenda before a call, generating a first outline for a report you’re already writing.

Open Claude or ChatGPT (either works here; most pro tiers start at $20/month, but the free version is fine to start). Give it real context about your actual situation, not a test scenario. Then evaluate what comes back.

The point is not to submit whatever the tool produces. The point is to notice how long it would have taken you to produce that same draft starting from scratch, versus how long it took you to edit what the AI gave you. That gap is your first win.

“All these companies are hyper competing right now… they all want your attention and they want people hyping them up. But what do I want these tools to do for me?” — Matt Hastings, in a Getting Started session with community members

That question is the right frame for phase one. Not “what can AI do?” but “what do I want it to do for me?”

What not to do in phase one

Don’t start with your most complex, high-stakes work. Don’t try to automate an entire workflow. Don’t spend time reading about AI instead of using it. AI is a practice skill, not a knowledge skill: the understanding comes through doing, not through preparation.

Don’t expect it to match your own output quality right away either. It needs training, just like you did! It needs to gain from your experience. You teach it to approach your skill level through repetition and refinement. It can help push your own skill level higher, but at the end of the day,

Phase 2: Build the Habit

Once you have one win, the work of phase two is to stop treating AI as something you “try” and start treating it as something you reach for. Regardless of whether AI can do the task from start to finish, bring it to the conversation anyway. Treat it like your new intern, don’t freeze it out of the interesting stuff.

This is harder than it sounds. New behaviors require repetition before they feel natural. Most people who abandon AI tools do it between weeks two and four, right after the novelty wears off and before the habit has set. Phase two is about surviving that gap.

The 20-minute daily practice

What works is small and consistent: 20 minutes a day, on real tasks, five days a week. Not a course. Not a weekend deep dive. Daily contact with the tool, using your actual work.

A few work styles to mix and match for AI experimentation:

1. Drafting: Take one thing you need to write this week and use Claude or ChatGPT to get a first draft. Edit it until it sounds like you.
2. Synthesizing: Paste a document you need to process (a report, a proposal, meeting notes) and ask a specific question about it. Not “summarize this.” Something like: “What are the three biggest risks in this proposal?”
3. Thought Partner: Bring a problem you’re stuck on to Claude. Describe it in detail. Ask for angles you might be missing. Push back on what it gives you, and make sure you’re not just looking for confirmation bias. I like to lie to it and frame my position as “a teammate says this, but I’m not sure. What do you think?” That way it resists the temptation to flatter me.

The variety matters. If you only ever use AI for drafting, you develop a narrow sense of what it can do. Rotating through different task types in phase two is how you start to see the fuller picture.

What you’re building in phase two

You’re building two things simultaneously: a habit (reaching for the tool without thinking about it) and a mental model (a feel for what kinds of tasks AI handles well versus where your own judgment is irreplaceable).

By the end of phase two, most people have 3-4 use cases that reliably save them time. They also have a growing sense of where AI falls short, which is just as valuable. Knowing when NOT to use the tool is part of getting good at the tool.

Phase 3: Expand Your Range, Build Something

Phase three is where the interesting stuff happens. By now you have the habit, you have the mental model, and you start to notice something: AI belongs in places you hadn’t considered.

This phase is less about following a plan and more about paying attention. When you hit a task that used to feel tedious, your brain now automatically asks: “Could Claude help with this?” That’s the shift you’re looking for.

You can also start to try building something from scratch, also known as Vibe Coding. An internal tool, a helpful new excel spreadsheet, a repeatable bank of prompts leading to a full workflow…try building something as a tool others can use without having to copy-paste your prompts.

What expands in phase three

Prompting gets better. You stop writing prompts that say “write me a summary” and start writing prompts that say “here is a 12-page proposal, here is my role on this project, and here are the three questions my director will ask in tomorrow’s review. What should I focus on?” The difference in output quality is significant, and you won’t understand it until you’ve lived it. You start using the “Prompt for Prompt” technique, realizing you can just ask the LLM for a better prompt for your main task before proceeding to the main task session itself!

Context management becomes a skill. You start giving AI background about who you are, your work context, your audience, and your constraints before asking for anything. This is the managerial relationship in practice: you’re managing inputs, goals, and output quality the way you’d manage a capable new hire who doesn’t know your world yet.

You start seeing building as possible. Some people in phase three start wondering if they could build something with AI, not just use it. A small internal tool. A custom workflow. A personal site. This is the point where vibe coding starts to feel like something worth trying, and that’s its own whole path.

The honest timeline

I want to say something real about timing, because a lot of AI content either overpromises (“transform your workflow in a week!”) or vaguely reassures (“just stick with it”). Neither is useful.

Here’s what the progression actually looks like for most people we coach:

Weeks 1-2: One or two wins. Still feels like effort. You’re choosing to use AI; it’s not automatic yet.

Weeks 3-6: A handful of reliable use cases. Reaching for the tool more naturally. Still getting surprised by what works and what doesn’t.

Month 2-3: AI feels like part of your toolkit rather than a separate project. You’re in conversations at work about how you’re using it. You have opinions about which tool to use for which tasks.

Month 4 and beyond: You stop thinking about “using AI.” You just work. Sometimes with AI, sometimes without it. The question is no longer “should I use AI for this?” It’s “what’s the fastest way to get a good result?”

That timeline is for people who practice consistently. Consistency is the lever. Someone who spends 20 minutes a day for six weeks will outpace someone who spends a full weekend every month.

A note on which tools to use

Phase one: Claude or ChatGPT. Either works. If you want one recommendation, Claude handles professional writing well and tends to follow complex instructions more carefully. Start there.

Phase two: Same tools, but consider the paid tier. Claude Pro and ChatGPT Plus are each $20/month. For the amount of time you’ll be saving, it’s worth it. You get better reasoning, longer conversations, and access to the best current models.

Phase three: Still Claude and ChatGPT as your core, but you’ll start developing opinions about when each is better for specific tasks. You may also start exploring tools that specialize in particular workflows, like Perplexity for research or Claude Code for building things.

Don’t try to evaluate every AI tool that launches. There will always be a new one, and most people who try to keep up with every announcement spend more time reading about AI than using it. Pick your core tools, get good at them, then expand.

The honest truth about the path

There is no shortcut from “I’ve barely tried this” to “I use AI every day and it’s changed how I work.” The path goes through actually doing it.

This is genuinely good news, by the way, not a warning. It means the advantage isn’t available to the person who reads the most or knows the most. It’s available to whoever shows up and practices. AI is for all of us, and the people getting the most from it right now are not the most technical. They’re the most persistent.

If you want to practice alongside other professionals working through the same phases, that’s exactly what happens in MVP Club’s weekly sessions. People share what they’re building, get unstuck, and pick up techniques from watching how others work. The progression goes faster when you’re not figuring it out alone.