Companies invest in artificial intelligence as if it were the magic solution, but the data reveals an uncomfortable reality: for every dollar spent, barely 18 cents generate real value. The rest is consumed in correcting errors, feeding useless data, or tasks no one asked for. At Amazon, employees resorted to AI to justify their work with trivial projects, inflating costs without tangible benefit. Citizens end up footing the bill without seeing improvements in services.
The hidden cost of automating the unnecessary 💸
From a technical standpoint, the problem is not AI, but its implementation without criteria. Language models like GPT or computer vision systems require clean data and clear objectives. If a company trains an algorithm to detect patterns in internal emails that contribute nothing, the result is a model that consumes GPU resources, electricity, and maintenance hours. Each bug fix costs more than what is saved. The key is to define return metrics before launching any machine learning project.
The AI they used to justify their morning coffee ☕
At Amazon, some teams created AI assistants for tasks like organizing office music playlists or reminding coworkers of birthdays. The result: spending on AWS servers that exceeded the salary of a human assistant. Meanwhile, customers were still waiting for delayed packages. The moral is simple: if your boss asks you for an AI project, make sure it's not just to make it look like you're doing something while he has another coffee.