Why the AI Budget Crisis Is an Engineering Problem, Not a Personnel One

The drive to fund artificial intelligence via mass layoffs has hit a wall. Gartner’s survey of 350 large-scale enterprises reveals that 80% of firms cutting staff to balance AI budgets see no corresponding improvement in returns. Uber serves as a cautionary tale: after granting 5,000 engineers AI coding tools, the company exhausted its entire 2026 AI budget by April. Despite 70% of code being AI-generated, COO Andrew Macdonald admitted the tangible impact on customer experience remains elusive.

Instead of treating payroll as the only flexible expense, companies are finding that token consumption is highly elastic when subjected to rigorous engineering. Security firm ProjectDiscovery, for instance, slashed its LLM spend by up to 70%—saving more than most layoff rounds—simply by optimizing prompt caching and restructuring workflows. By shifting from expensive flagship models to right-sized tiers, utilizing batch processing, and implementing retrieval-augmented generation, firms can achieve significant savings without shedding talent.

The Human Dividend

Optimizing compute costs is only half the equation; the other is retaining the expertise required to manage these systems. Klarna’s experiment with replacing 700 customer service agents with an AI assistant resulted in a decline in quality, forcing a pivot to a blended model. As Gartner predicts that half of companies that cut service staff will be forced to rehire by 2027, the urgency of preserving junior developer pipelines becomes clear. Stanford’s Institute for Human-Centered AI notes a 20% drop in employment for entry-level developers, a trend that threatens the future supply of senior engineers. The most successful organizations are those that treat the token budget as a variable to be engineered, using the resulting savings to invest in the human judgment necessary to make AI output valuable.