
There was a week, I cannot tell you exactly when it happened, when I was technically everywhere. Responding within minutes. Never missing a message. Always available. And yet, two people I genuinely care about told me, in different ways and at different moments, that something felt off. One was a colleague. One was a friend. Neither of them could fully explain it. They just said our conversations had started to feel different. I knew what it was. It was not me. I run a company that helps brands become visible to AI systems . I believe in automation. I use agents every day for research, content distribution, follow-up sequences, and inbox management. I am not writing this to argue that AI is bad. I am writing this because I came uncomfortably close to automating myself out of my own relationships, and I almost did not notice. Here is what actually happens when you delegate communication to agents. The system works . Genuinely well. Often better than you do. Messages go out on time. Nothing falls through the cracks. Response rates improve. People begin to experience you as more efficient, more reliable, more organized. But there is a layer underneath all of that that the metrics never capture. The person on the other side starts to feel it. Not immediately. Not dramatically. Just gradually, quietly, they stop expecting anything from you that feels personal — because nothing personal is coming. I felt this from the other side too. A few months ago, I was trying to order fabric from a supplier I had worked with before. Something about the material required a real conversation. Not a ticket. Not a dropdown menu. A person who understood what I was trying to make. The AI handled every step perfectly . It was fast, polite, and technically correct. And I felt completely alone in that interaction. I left. Not because the process failed. Because I was being processed. There is a difference. And people feel it. Klarna ran into a version of this problem at a scale large enough to make headlines. In early 2024 , the company announced that its AI assistant had handled 2.3 million conversations in its first month across 23 markets — work it said was equivalent to 700 full-time agents. Response times dropped dramatically. On paper, the system looked like a success. But by 2025, Klarna’s CEO was publicly acknowledging something the original efficiency metrics had failed to capture: cost had become too dominant a lens, and quality had suffered. The company began moving back toward a more human-supported model. Not because the technology failed. Because the technology succeeded at the wrong thing. That distinction matters more than most companies realize. People consistently say they value empathy and human connection more than pure response speed when something actually matters. That should not be surprising. What surprises me is how often founders, operators, and leaders assume that principle applies only to customer support — and not to the rest of communication. I think it applies everywhere. To colleagues. \n To clients. \n To new relationships. \n To difficult conversations. \n To trust. This is the part of agentic systems that I think we still do not talk about honestly enough. When you remove yourself from the small moments — the reply that took a beat longer because you actually thought about it, the message that was slightly imperfect because a real person wrote it in a real moment, you also remove some of the signals that tell another person they matter. AI removes the imperfection And sometimes the imperfection was the only evidence of presence. I am not saying we should stop using agents. I use them every day, and I am not going back. But I have started asking a different question before I automate anything: Does the person on the other side of this need to feel that I showed up? Not: can this be automated? \n Not: can this be handled faster? \n Not: can this be routed more cleanly? But: does this moment require presence? If the answer is yes, I show up. Not because it scales. Because some things should still be allowed to matter more than scale \ \
View original source — Hacker Noon ↗



