What I’ve Learned Building with AI Agents (And the Tools I Actually Use)

Over the past year, I’ve been hands-on with AI agents and agentic automation—experimenting with everything from automating research tasks to building digital “teammates” that can collaborate on projects. My journey started with simple automations, often just connecting a few apps to save time on repetitive work. I quickly realized that you don’t need to build something perfect from the start. In fact, my earliest agents were far from flawless, but each iteration taught me something new and pushed me to keep improving.

One of the biggest lessons I’ve learned is the value of starting small and separating your agents by function. It’s tempting to try to create a single “super agent” that does everything, but I’ve found it much more effective to build specialized agents for specific tasks and then have them interact. Think of it like assembling a team—each agent has a clear role, and together they accomplish more than any one could alone.

Even though many of these tools are marketed as “no-code,” building effective AI agents is more complex than it looks. Success depends on how clearly you can define instructions and workflows. Prompt engineering—crafting precise, detailed instructions for your agents—is absolutely essential. I’ve learned to treat my agents like interns: the more guidance and feedback I give, the better they perform.

The pace of AI development is relentless, and it’s easy to feel like you’re always behind. My advice is to just start. The sooner you get your hands dirty, the more prepared you’ll be as the technology evolves. Along the way, I’ve run into plenty of unexpected challenges—debugging agent workflows can be tricky, and sometimes agents make decisions you didn’t anticipate, especially if your prompts are vague. But the process of iterating, testing, and learning from mistakes has been invaluable.

For anyone looking to get started, here are the tools I’ve found most useful:

• n8n: My go-to for flexible, open-source workflow automation. It’s great for connecting different apps and building custom automations.

• Zapier (Zap Agents): Super user-friendly for linking apps and automating tasks without much setup.

• make.com: A visual platform that makes it easy to design and manage more complex workflows.

I’m still learning, and there’s a lot I don’t know—but that’s part of what makes this space so exciting. If you’re experimenting with AI agents or have your own favorite tools, I’d love to connect and swap notes. The best time to start is now.