I'm not a career coder. My background is in business analysis and project management, not computer science. But over the past few years, I've gone from writing SQL queries to building agentic AI systems. Here's how that happened.

The Accidental Start

It started with curiosity. I was managing AI/ML projects, translating between engineers and stakeholders, and I kept wondering: what's actually happening under the hood? I could explain what a model did at a high level, but I wanted to understand how.

So I started small. Python tutorials. Pandas for data analysis. Basic scikit-learn models. Nothing fancy—just enough to follow along when engineers talked about training data and feature engineering.

The Turning Point

The real shift came when I stopped just learning and started building. I had a problem at work: our team was manually tracking project updates across multiple tools, and things kept falling through the cracks. I thought, "could I automate this?"

My first attempt was terrible. Buggy code, fragile integrations, barely worked. But it worked enough to be useful. And more importantly, I learned more from building that one thing than from months of tutorials.

Learning by Prototyping

That became my approach: learn by building things that solve real problems. Each project taught me something new. An API integration here, a data pipeline there, gradually more complex systems.

When LLMs exploded onto the scene, I was fascinated. Not because of the hype, but because they solved a problem I'd always struggled with: making AI systems that could handle unstructured, messy, real-world data. Suddenly, prototypes that would've taken weeks could be built in days.

The AI Exploration Phase

Now I'm deep into prompt engineering, RAG systems, and agentic frameworks. Building tools that use LLMs to automate workflows, analyze data, and assist with decision-making. It's not about being a "real programmer"—it's about using these tools to solve problems.

The business analysis background actually helps. I think about user needs, edge cases, what happens when things break. I focus on building systems that people will actually use, not just impressive demos.

Advice for Other Non-Coders

If you're in a similar position—working with AI but not from a coding background—here's what worked for me:

Start with a real problem. Don't just do tutorials. Build something you actually need, even if it's small and messy.

Embrace being a beginner. Your code won't be perfect. That's fine. Make it work first, make it better later.

Use your domain expertise. You understand the business context, the user needs, the real-world constraints. That's incredibly valuable when building AI systems.

The AI field is moving so fast that everyone's learning constantly. Being a "non-coder" matters less than being curious, persistent, and focused on solving real problems. Just start building.