I’m a Working Dad Who Built Over 150 AI Agents to Run My Life | Invesloan.com

This as-told-to essay is based on a conversation with Tristen O’Brien, a 44-year-old AI educator and content creator who works in cybersecurity based in Indiana. His words have been edited for length and clarity.

For a long time, I’ve felt like I can only focus on three things at a time without everything falling apart: family, work, and school. If I tried to add something else, one of those three things would fail.

I’ve been in the tech industry for over 20 years, and I’m working toward my doctorate in AI. I’m also a YouTuber. I spent years learning the programming language Python because I wanted to create a world where I could automate parts of my life, giving me more time for hobbies and other things.

Then ChatGPT came out, and I realized I had wasted about five years doing that.

Since learning how to build AI agents, I’ve built over 150 of them to help me run my life. I’m now down to using around 70.

I’ve always wanted to automate parts of my life

There’s a book about automating the boring parts of your life with the programming language Python. Years ago, I found this book and decided to learn it. Then, when ChatGPT came out, I quickly got into it.

I’m not a developer; I do this by vibe coding. I first used ChatGPT to make funny songs and things like that. My interest kept growing, and I kept doing more with it. I’ve switched to Claude as my main model now, but I still use ChatGPT, Perplexity, and Gemini.

Two years ago, I went to an AI conference where I learned about AI agents. I remember being so excited about what I had learned that I told myself that as soon as I got home, I was finally going to automate my life with AI.

I use a specific prompting technique to build my agents

The very first agent that I built after the conference was a Chipotle burrito agent that could order my burrito for me. I love Chipotle, and it was silly, but that agent helped me see the power of AI.

When I build agents, I’m never just like, “OK, Claude, this is what I want you to do, do the whole thing.” Claude has a command called “/goal” that keeps the agent going until it gets the goal right. I think it’s helpful, but I’ve learned to be very clear, or it will mess up.

I do what I call my BIP process: Braindump, interview, prompt. I brain-dump everything, explaining what I want it to do, then I have Claude ask me questions about anything unclear or that it thinks I didn’t ask. Then it gives me this crazy long prompt. So I might’ve started with a 15-word prompt, but now I have a 500-word prompt on everything that I want the agent to do.

I’ve built over 100 agents and stopped using some of them

In total, I’ve built five research and education agents, four research and news agents, one agent that builds new products, 18 financial agents, 27 home and family agents, 26 inbox and communication agents, 42 agents for my YouTube channel, and another 27 agents for publicity and outreach.

Not all of those agents run at once. For example, I only have 12 active home and family agents: Four run every day for reminders about birthdays, softball, the dog, and school paperwork. Four run weekly for groceries, meal planning, a Sunday family brief, and a weekly check-in coach. One runs monthly for house maintenance reminders, and three run only when something comes up, like travel, packing, or gift ideas.

I also had an agent who looked at a Google Sheet with all of my friends’ and family’s birthdays. If it were two weeks before their birthday, the agent would go to a webpage for a service that would send them a birthday card with what looks like my handwriting, then postmark it from my address. But I stopped using that one because it didn’t feel personal enough.

AI is more powerful once you learn how to use plug-ins

I have a wife and two kids. I think my daughter likes AI a lot, too, because we talk about it a lot, and my wife sometimes gets into it. But my son is an artist, so he’s not as big a fan, and I agree that there’s definitely a line with creative work and industries.

However, one of my favorite agents I’ve built is our family AI bot, which we all use.

I used the texting service Twilio with a Claude plug-in to create an AI agent that is a custom chatbot and assistant for our family. It pings us, and we’re able to vote on things like what we want for dinner that week, or see our calendars side by side. This agent helps us organize our lives and stay on the same page.

AI gets really powerful when you start using connectors. My Claude agents wouldn’t be as powerful if it couldn’t connect to my Gmail, my calendar, or Twilio. So, from a technical standpoint, I’d suggest learning how to sign in on those kinds of plug-ins first.

I spend around $600 a month on my AI subscriptions, which also support my YouTube channel and other work. If I were just using this for my personal life, I’d probably scale back to the $200 Claude Max subscription.

I created an agent hierarchy game

I get digital overload sometimes, and there is a point where too much is digital. It was hard to keep up with all my agents, so, in addition to getting rid of some of them, I had to create a digital hierarchy.

I’m the chief human officer. I have my AI chief of staff, who works with a handful of director-level AI agents, and those directors each have AI worker agents under them. My chief of staff will relay updates from other agents, so now I’m mainly communicating with one agent who’s relaying updates to the rest, which helps with the overload.

I also created an animated game world where I can see all my agents running around a village. The different levels of agents have different outfits, so I can tell which one’s which. I can run around the game world with my chief of staff and go talk to one of my agents any time to check in on what it’s doing.

I can also press a button to call a meeting, and we all go to this meeting hall in the game. It’s more of a visual way to see all of my agents, and talk to them individually if I want to.

I worry about privacy, but I’m getting more time back

There are a lot of things I don’t share with AI, like my Social Security number, but you have to have a level of trust, as you do with Gmail or any other service that is using your data or has access to it.

I feel like AI does more good than harm. I’m not a doomsday guy about AI; I’m more of a utopian. There are things that I’ve picked up because I have more free time. I like to 3D print, fly drones, and ride my bike. I feel like I get a lot more done now.

That was my mission from the start. To give myself back some time to do things that I really enjoy and appreciate, while still being able to focus on the top three things that matter to me most.

Do you have a story to share about how you’re using AI agents? If so, please reach out to the reporter at [email protected].