You just hired a million bad employees': How the brief tokenmaxxing era delivered the opposite of what it promised | Fortune

The tokenmaxxing era delivered the opposite of what it promised focuses on concerns that AI agents are being deployed faster than companies can manage them. George Sivulka, founder of the AI enterprise startup Hebbia, says enterprises raced to roll out AI workforces without building the management infrastructure to run them. In an essay referenced by Fortune, he argues AI did not cut labor costs and instead reversed the relationship, stating that “for the first time in history, humans are cheaper than software.” He warns that poorly specified tasks cause agents to fail, estimating that only 1 in 100 employees can provide enough context for agents. The article links such failures to “looping,” where agents repeatedly call themselves to self-correct, described as “spending tokens on spending tokens.” It also cites how losses were disclosed by Amazon and a shift by Ford toward human engineers alongside AI.