#63 · AI Automation: Too Good To Be True?
I’m not very familiar with building agents in Python or LangChain, but most modern agentic flows are now built in platforms like NAN. In my view agents aren’t a good solution for many business use cases, so I wouldn’t recommend them.
I think the agent‑agent protocol is pretty cool, but I don’t think it makes too much sense to use an agent multi‑agent framework right now. There’s a lot of early hype with these technologies, followed by a crest/valley of disappointment, and after 6‑12 months the tech gets good—we’re at that point with multi‑agent frameworks. I don’t think the tech is anywhere near as good as building a simple traditional workflow, though I’ve experimented with both platforms and like them; they’re cool.
My recommendation is to start with one of my NAND courses to get up to speed on visual processes and how automation works in those flows. Immediately after, take a big cloud‑code course to learn how to apply Claude to that foundation. It's important to understand basics like JSON and how data moves through nodes on a graph; once you have that, Claude code becomes capable of handling a lot for you.
The AI agent I use most is Claude, specifically Claude code (cloud code). It handles most of what I need, even though it can be verbose and sometimes requires scrolling through output. I connect Claude to various platforms via built‑in plugins or connectors, give it access to my email and other tools, and use remote control to have a virtual AI assistant on my phone, computer, etc. I also use Codex and occasionally other tools, but Claude remains my primary assistant.
First, I’d join a community like Maker School—not just to use my own products but because membership gives you deep discounts (often a 10x return) on AI platform credits and payment‑processing fees. With those savings I’d start by learning a visual drag‑and‑drop automation builder (such as the n platform) for about four to six hours; that teaches you the fundamentals of systems, workflows, JSON, HTTP requests, etc., which accelerates everything else. Next I’d spend another four hours on a Claude‑code course that covers the technical setup, agent communication, model selection, and so on—roughly ten hours total. After that foundation, I’d devote virtually all remaining time to acquiring customers; the first 90 days would be less than 1% spent on learning and the rest on outreach. You need a customer to actually enter the arena, so I’d use Maker School’s day‑by‑day accountability to hit call and outreach targets, then loop back to upsell from any lower tier to higher ones as the relationship grows.