For the last two years, I’ve been on a wild ride through the world of AI. It started innocently enough, with a single bookmark in my browser. That bookmark became a folder, and that folder became a sprawling collection of tools that I’ve tinkered with, tested, and sometimes, been completely amazed by. I’ve chatted with ChatGPT, researched with Perplexity, created with Midjourney, and explored with Gemini. But the one that truly stuck with me, the one that felt different from the start, was Manus.im.

This past week, everything changed. I built my own AI agent. And let me tell you, it was an experience. Not because of the coding or the technical stuff—AI helped me with that, of course. The real work, the hard work, was figuring out what I wanted it to be. How should it think? What should it care about? What makes a piece of information valuable to me? I spent days, on and off, just sitting with those questions. It was a journey into my own mind as much as it was a journey into AI. The result is an agent that scours the world for news about AI and education, but it does it with my perspective, my curiosity. It’s not just a tool; it feels like a part of me.

For two years, I’ve felt like I’ve been doing the work of five people, thanks to AI. But this is different. This is a game-changer. An AI agent isn’t just about getting a quick answer. It’s about having a partner that knows how you think, that gets your style, that understands what you mean even when you don’t say it perfectly. It’s a conversation, not a command.

And that’s what makes me a little scared, or a little bit more.

Agentic AI is the part of this whole AI revolution that keeps me up at night. I’ve heard all the arguments about superintelligence and consciousness, and honestly, I don’t think we’re anywhere close to making a machine with a soul. There’s a spark in the human heart that I don’t think we’ll ever be able to code.

But agentic AI isn’t trying to be human. It’s trying to do a human’s job. And it’s already happening. I heard about a big company that’s letting people go, but not in some big, splashy layoff. They’re doing it quietly, five or ten people at a time, and replacing them with AI agents. It’s silent. It’s efficient. And it’s happening now.

I got into a debate on social media the other day with a programmer who was adamant that AI wouldn’t take his job. But then he admitted that his company hasn’t had to hire any new programmers because he’s so much more productive with AI. He didn’t even see the irony. He was making my point for me.

As we get better and better at creating these tireless, soulless agents, I think we need to have a real, human conversation. Every company, every university, every one of us. We need to ask ourselves: what are the things we should never outsource to a machine? What are the human jobs that are about more than just function?

Which brings me to Krispy Kreme

I went with my kids to Krispy Kreme the other day, a beautiful, chaotic mess of four children under ten in our car. And yes, it was just for them. Of course, it was solely for their benefit. We pulled up to the drive-thru, and a cheerful, slightly-too-perfect voice greeted us. “Hi, welcome to Krispy Kreme?” It became apparent very quickly that it was an automated/AI machine getting ready to take our order.

What happened next was pure, unadulterated family life. A flurry of kids' voices, all talking at once, a tidal wave of donut demands, second thoughts, and sibling rivalries. I was just trying to keep my head above water, trying to translate the beautiful chaos into a coherent order. And as I looked at the screen, I saw the AI diligently and carelessly capturing every single word. Before I could even open my mouth to try and explain, the AI voice cut through the noise: “I captured all that. You have 73 donuts. Please pull forward to the window.”

For a split second, I panicked. 73 donuts? My car, my wallet, my sanity… none of them were prepared for that. And then, a real, human voice broke through. “Sorry about that,” she said, with a warmth that the AI could never fake. “Don’t worry about that glitch. How can I help you?”

That question. That simple, human question. It was everything.

This experience left an impression on me. We’re going to be hearing fewer and fewer human voices asking that question. And while I’m all for progress and efficiency, I think we need to be very, very careful that we don’t replace the soul behind the help.

There are some things a machine can’t do. And as agentic AI accelerates and these systems become more capable and invisible in the background of our lives, we need to name those things with clarity and conviction.

The next frontier is not just technological. It is human. It is about choosing which spaces must remain unmistakably ours. These are spaces where the importance of presence is paramount. Where empathy matters. Where the tone of an authentic voice can calm an overwhelmed parent in a drive-thru. Where meaning is shaped not by efficiency but by connection.

AI will keep getting better. It will help us do more, move faster, and stretch ourselves beyond what we thought possible. But if we are not intentional, we may wake up one day to find we have outsourced not just tasks, but touch. Not just work, but the warmth that makes work worth doing.

So yes, build the agents. Explore the tools. Strive to exceed the limits. But hold tight to the question that only a human can ask with a soul behind it.

How can I help you?

Because in the age of agentic AI, that question might just become our most important act of resistance, our most important act of leadership, and our most important act of being human.