I’ve spent almost 2 years experimenting with AI, mostly failing. Not because AI doesn’t work, but because I didn’t understand how to use it. I looked for one solution that would handle everything—a universal assistant that understood my style, patterns, workflows and habits all at once. None of it worked.
Think about it for a second. How is an AI supposed to learn what “good” looks like when you’re throwing ten different problems at it simultaneously? It’s trying to optimize for everything, which means it’s optimizing nothing.
Then I changed my approach and focused on one small problem at a time. One specific process. And suddenly, things started actually working.
The best example? Specialized assistants. Email assistant handles emails. Calendar assistant manages scheduling. Meeting notes assistant captures discussions, extracts key points and stores them in Obsidian.
Here’s what surprised me most about this approach. Each specialized assistant got really good really fast because it had clear, focused context. It knew exactly what I needed from it, no confusion about priorities, not trying to balance competing demands. Just one job. And once one assistant works, you naturally build the next one, and the next. They start working together because each one clearly knows its role.
Small, specialized tools aren’t a compromise or a temporary solution. They’re how you actually learn what AI can do for your business.
Start small, ship fast, learn what works. Throw away what doesn’t.
Are you planning or are you shipping?