Stop Studying It. Go Build Something.

Individual Contributor Series, Article 3 of 3

This is the article where we stop talking about it.

The first article was about why the window is open and why waiting is the wrong move. The second was about the foundation you already have and how to find your starting point. This one is the starting gun.

Here's exactly what to do, what it costs, and what you get on the other side.

The Process That Never Actually Changed

Before we talk about AI tools, let's talk about something that hasn't changed in thirty years.

Software Development Life Cycle. SDLC. The foundational process for how you take an idea and turn it into something that actually works.

Requirements. Design. Technical design. Build. Test. Deploy. Iterate.

That's it. That's been the framework for building solutions since before most of the current AI tools existed. Since before the internet was mainstream. Since before anyone was arguing about certifications for technologies that didn't exist yet.

If you've been part of any project — any process improvement, any system implementation, any operational change — you already understand the bones of this. You've lived in some version of it. You may not have called it SDLC but you've done it.

Start here. Before you open a single AI tool.

What are the requirements for what you're trying to build? Not a vague goal — specific requirements. What does it need to do? Who uses it? What does good output look like? What are the constraints?

What does the design look like? Not the technical design yet — the conceptual one. If you close your eyes and imagine the finished thing working the way you want it to work, what are you seeing? What does someone do with it? What do they get out of it?

Then the technical design. How does the information flow? What goes in? What comes out? What happens in between? Where are the decision points?

Then the workflows. Step by step. What triggers the process? What happens next? And after that? Where does it end and what does done look like?

This is how every good solution gets built. It's how a two-person startup builds software. It's how a Fortune 100 enterprise runs a program. The vocabulary might be fancier in the enterprise version but the logic is identical.

Write this down before you build anything. Your requirements. Your design. Your workflows. Even rough. Even imperfect.

That document becomes the instruction set you hand to AI. And AI with a clear instruction set produces dramatically better results than AI with a vague one.

Use AI to Learn AI

Here's the shortcut that nobody in the certification industry wants to advertise.

The most efficient way to learn AI is to use AI to teach you.

Not passively. Actively. Specifically.

Ask it what the difference is between different models and what each one is better at. Ask it how to write a better prompt for the specific thing you're trying to do. Ask it why the output it gave you isn't quite right and what you should tell it differently. Ask it to explain a concept you don't understand in plain language, then ask it to explain it a different way if the first explanation didn't land.

AI is infinitely patient. It doesn't get tired of your questions. It doesn't judge you for not knowing something. It doesn't move on before you're ready.

Use it the way you'd use the most knowledgeable, most available mentor you've ever had — except this one is available at midnight when you're working on your own time and doesn't have a billing rate.

Ask it how to keep it focused on what you're trying to build. Ask it what information it needs from you to give you better answers. Ask it to review what you've built and tell you what's missing or what could break.

The learning curve people spend twelve months on in a certification program compresses dramatically when you're applying the learning to a real problem in real time instead of studying concepts in the abstract.

The Right Tool for the Right Job

Here's something the "just use ChatGPT" crowd skips over.

Different AI models are built differently and are genuinely better at different things. Using the right tool for the right job isn't advanced behavior — it's basic craft. And once you know roughly what each one is good at, switching between them becomes second nature.

A few examples to get you oriented.

Perplexity is built for research. If you need to understand a topic, find current information, or survey what's out there on a subject, Perplexity pulls from live sources and cites them. It's the research assistant you use before you start building — and while you're building, when you hit something you don't understand.

Grok has a different character — built with a broad training base that makes it useful for things like understanding customer sentiment, reviewing code from a different angle, or getting a take that cuts through conventional thinking. Worth having in the rotation for a second opinion.

Claude and OpenAI's models are strong at reasoning, writing, and code generation. For building — actually writing functional code, structuring logic, drafting documentation — these are workhorses. Different people have preferences. Try both. You'll develop your own.

Image and design models exist specifically for visual work — generating UI mockups, creating assets, producing visual concepts faster than any traditional design process. If your build has a visual component, these belong in the toolkit.

You don't need all of them on day one. Start with one or two. Add as the problem requires it. The point is to stop thinking of "AI" as a single thing and start thinking of it as a toolkit where different tools have different strengths — the same way you'd think about any other set of professional tools.

What This Actually Costs

Let's put real numbers on this.

Most of the tools described above have free tiers that are genuinely useful for getting started. The paid tiers — the ones that unlock higher usage limits and more capable models — run roughly $20 per month per tool.

Pick two or three tools. Maybe $40 to $60 a month total. Call it $100 at the high end if you add a specialized tool for your specific use case.

Spend 90 days building something real with that investment.

Compare that to a certification program. Most of them run several hundred to several thousand dollars. They take months. They teach concepts. They give you a credential that signals you studied.

The $100 over 90 days gives you something different. It gives you a thing you built. A problem you solved. A process you documented. Results you can point to. And an honest answer to the question of whether this is something you want to keep doing.

At the end of 90 days, one of three things will be true.

You'll love it. You'll discover that you have a genuine talent and passion for this kind of work — that the combination of your business knowledge and these tools is something you want to go deep on. That realization changes what you do next and what you're worth in the market.

You'll like it. You'll be more effective at your current job because you understand these tools well enough to use them well. You won't be building AI systems full-time but you'll be better at everything you already do. That has value too.

Or you'll decide it's not your passion. And that's a completely legitimate outcome. You'll know that from experience instead of assumption. And you'll understand what's actually in the mystery black box — which means you'll never again sit in a meeting about AI without knowing whether what someone is describing is real or theater.

Any of those three outcomes is worth $100 a month and 90 days.

Fair Warning: This Gets Addictive

Here's something the tidy "just try it" advice usually leaves out.

For people who are wired to solve problems and build things, this doesn't stay at $100 a month forever.

At some point you'll hit a usage ceiling on a $20 subscription right in the middle of something you're building. You'll have to make a choice: slow down and wait for the monthly reset, or move up to the next tier. For people who are genuinely lit up by what they're building, the choice is usually obvious.

Then the build itself gets bigger. You start thinking about deployment — where does this thing actually live so other people can use it? You start thinking about data — where does the information get stored and how does it move? You start thinking about automation — what if this process ran itself instead of requiring you to trigger it manually?

That's when tools like Vercel, Supabase, Cursor, and n8n start showing up in your browser tabs. Deployment infrastructure. Databases. AI-assisted coding environments. Workflow automation. Each one adds capability. Each one adds a little to the monthly cost.

You choose what to add based on what you're building and what makes sense for your budget. Nobody is telling you what to spend. This is your experiment.

But here's the reframe that matters.

When you're building something at that scale — something with real infrastructure, real functionality, something that solves a real problem for real users — you're not running a hobby. You're building a product. And the tools you're paying for are business expenses for software that could generate revenue if you choose to sell it, license it, or use it as the foundation of something larger.

These skills don't belong to your employer. They belong to you. You take them with you. You apply them anywhere. You build with them on your own time toward your own outcomes.

The market for people who can actually do this — not describe it, not certify it, but build real things that work — is not shrinking.

Find Your People

You don't have to figure this out alone.

There are communities of people doing exactly what this article is describing — learning by building, sharing what they're working on, asking questions, helping each other debug problems and think through designs. Some of them are on Reddit. Some are in Discord servers organized around specific tools. Some are in LinkedIn groups. Some are local meetups.

These communities are not full of credentialed experts performing expertise at each other. They're full of practitioners at different stages of the same journey — people who picked up a problem, started building, hit a wall, asked for help, got unstuck, and kept going.

Find one that feels like your people. Show up. Ask questions. Share what you're working on. Help someone who's a few steps behind where you are.

The isolation of learning something new on your own is one of the biggest reasons people quit before they get traction. The community is the antidote.

The Only Permission That Matters

Thirty years from now, someone is going to write an article about all the job titles that exist in 2055 that nobody could have predicted today. They'll talk about how the roles evolved from the current generation of AI tools the way today's roles evolved from mainframes and early software.

They'll describe the people who got there first. The ones who moved when the window was open instead of waiting for someone to hand them a roadmap.

Those people are building right now. Nights and weekends. Lunch breaks. After the kids go to bed. With $60 worth of subscriptions and a real problem they wanted to solve and the same foundational skills they've been building their whole career.

They're not waiting for a certification. They're not waiting for their manager to create a development plan. They're not waiting for the company AI strategy to be finalized.

They're building. And when the window fully opens — when every organization is actively looking for people who can actually do this work — they'll have a year of real experience and real output and real results that no standardized test can replicate.

That can be you.

Not someday. Not after the next performance review. Not after the certification program that starts in the fall.

This month. With tools you can access today. On a problem you already understand better than any outside consultant ever will.

AI is not a spectator sport.

And the only permission you need to start learning something new?

That belongs to you.

That's the Individual Contributor Series. Six articles total across both series — written for the leaders who need to change how they're thinking and the practitioners who are done waiting for those leaders to catch up.

If your organization is ready to stop theorizing and start building — 2weekAI delivers fixed-bid AI assessments and deployments in two weeks. Real pilots. Real results. Discovery call to proposal in 48 hours.