The Trend of Firing Humans for AI… and the Reality
In recent years, some companies have jumped headfirst into replacing employees with artificial intelligence. The promise was clear: lower costs, more speed, and almost magical efficiency. But of course, reality doesn't always render as expected. Many of those companies that dreamed of offices full of AI agents solving tasks as if they were automation scripts in Houdini… are now realizing that the human touch is not so easy to replicate.
The Klarna Case: From Bragging to Backpedaling
One of the most talked-about examples is Klarna. Its CEO publicly boasted that an AI was doing the work of 700 fired employees. But a year later, customers began complaining about the service quality, and the company had to hire human staff again. The result was a customer service workflow that seemed more like a poorly resolved logic bug in Unreal Engine than a functional service.
AI Yes… But Closely Supervised
According to a Gartner survey, half of the companies that started this process have backed out. Most are opting for a hybrid system, combining AI and humans. It's not that AI doesn't work, but leaving it alone to manage customers is like doing a lighting bake without checking the normals first: disaster is almost certain.
Experiments That End in Glitches
An experiment from Carnegie Mellon University makes it clear: they put AI agents in charge of a fictional company, assigning them roles in finance, administration, or software engineering. The result: they only completed 24% of the tasks. In other words, not even with a post-production pass in After Effects could that project be saved.
Not All Failure, But No Miracle Either
Yes, there are companies where the change worked out well. Startups that cut staff and improved results thanks to AI. Even Duolingo reduced its team of translators without the service collapsing. But for every success story, there are others like UPS or Cisco that had to rethink the strategy. IBM, for example, has had to hire more staff just to supervise its AI systems. Something like having to buy more RAM to use Arnold's denoise without the PC freezing.
Less Hype and More Stress Tests
AI promises a lot, but it still has to go through many beta phases before truly replacing humans. If you're going to bet on it, better as part of the pipeline, not as the only actor in the project. Otherwise, the crashes come… and we're not talking about After Effects precisely 💥.