When I tell people I build agentic AI systems, they assume I have a computer science background. I studied immunology. The gap between those two things is simultaneously wider and narrower than you’d think.
What you actually need to know
Building AI agents isn’t about understanding backpropagation. It’s about:
- Clear problem decomposition: breaking a complex task into steps an AI can execute
- Prompt design: communicating intent precisely (turns out, scientific writing is great training for this)
- System thinking: understanding how autonomous components interact, fail, and recover
- Judgment calls: knowing when to automate and when to keep a human in the loop
The tool that changed everything
Claude Code is what made this possible for me. It gives you a conversational interface to build software. You describe what you want, it writes the code, you review and iterate. The key insight is that you don’t need to write code yourself. You need to think clearly about what the system should do.
Real examples
A voice-activated expense tracker. I connected Telegram, ChatGPT, Make.com, and Google Sheets so I could speak into my phone and have expenses automatically logged and categorized. I configured every automation step manually.
This portfolio website. Built from scratch with Claude Code. I described the layout, the interactions, the dark theme. Claude handled the Astro framework and CSS. I made every design and content decision.
A career opportunity scanner. It monitors company career pages via their APIs, scores each position against my profile, tailors resumes to match the job description, and generates application materials in batches. The whole system runs through Claude Code skills and scripts.
Each of these started the same way: I had a clear problem, broke it into steps, and used AI to handle the implementation I couldn’t do myself.
The real skill
The people who get the most out of AI tools aren’t the best programmers. They’re the best problem decomposers. If you can take a messy, real-world problem and turn it into a sequence of clear, testable steps, you can build almost anything with the tools that exist today.
That’s a skill that transfers from any analytical background. Science, finance, operations, consulting. If you’ve ever designed an experiment or built a process, you already have the foundation.