A Reddit discussion about people losing their trust in AI agents. It hit close to home. Because what I've been seeing in the field confirms every word of it.
I have been working in tech for long enough to recognize bullshit when I see it. And right now? The AI agent space is completely full of it.
Every single week there is a new startup claiming their agent can "autonomously manage your complete business workflow." Really? Because I tried one last month that could not even send a simple email without messing up the recipient list. Great experience.
"The worst part is not the technical limitations. It is the dishonesty around them."
Companies know their agents are failing constantly — and they simply don't mention it. Not at all. That is not a product limitation. That is a trust problem.
In a recent presentation, Karpathy said what many of us have been thinking: the fully autonomous agent is mostly fantasy right now. We should be building tools that help humans — not trying to replace them entirely.
Follow on X →He is right. But that does not sell well to investors, does it?
Here is what actually bothers me about the current state of AI agents:
This is a pattern we have seen before.
Everything was going to be "revolutionized." Supply chains, healthcare, voting, banking — all of it. How did that work out for most companies?
Everything is going to be "autonomously managed." Operations, workflows, customer service, content — all of it. We are watching the same movie again.
Maybe there are some decent agents out there doing genuinely useful work. But the signal-to-noise ratio is terrible right now. Too much hype, not enough honesty about what is actually working.
I had to explain to my team last week why the "revolutionary AI agent" we tried implementing for our own content workflow actually created more problems than it solved. That was an interesting conversation. And we are a company that runs AI across 7+ brands every day — so if it does not work for us, it is probably not working for most.
Maybe this is a German thing. We have this tendency to be skeptical of overpromising. It probably comes from an engineering background — we like things that actually work, not just things that sound impressive in a PowerPoint.
The autonomous agent thing has gotten completely out of hand. And the people quietly building reliable, human-in-the-loop AI tools? They are going to win in the end.
Is anyone else experiencing this fatigue — or am I just becoming old and grumpy about new technologies?