A large company like Uber has acknowledged that it cannot demonstrate that its spending on artificial intelligence improves business results. Although almost all of its engineers use AI and much of the code is generated with assistance, the annual budget was exhausted in just four months. This reflects a real problem: AI costs money that is easily measured, but its benefits are difficult to prove to the public.
The hidden cost of automating with AI 💸
Development teams integrate code assistants and generative models to speed up tasks, but the time savings do not always translate into revenue. Uber spent millions on cloud infrastructure and model licenses without a clear return. The paradox is that, although AI-produced code reduces minor errors, operational costs skyrocket. For companies, it is a gamble without guarantee: the expense is concrete, the profit, an assumption.
AI eats the budget and gives no receipt 🧾
It turns out that artificial intelligence is like that friend who orders pizza for everyone but never chips in. Uber swiped the credit card in the AI machine and within four months was in the red, unable to show a single report saying: look, this has made us rich. Meanwhile, engineers keep asking AI to write the code for them, but the CFO looks at the bill and wonders if it wasn't cheaper to go back to typing by hand.