Companies have deployed AI agents too quickly without the necessary infrastructure to control their operations and costs. This conclusion was drawn by Hebbia founder George Sivulka in an essay for the a16z newsletter, as reported by Fortune.
He linked the issue to tokenmaxxing—a practice where organizations tracked and encouraged the growth of tokens consumption in AI operations. According to him, companies mistakenly viewed increased token consumption as a sign of AI adoption and productivity improvement.
“You’ve just hired a million bad employees,” Sivulka wrote.
In his view, AI agents have not so much reduced labor costs as they have altered the cost structure. In the essay, Sivulka stated that “for the first time in history, people have become cheaper than software.”
Hebbia develops corporate AI tools. According to Fortune, the startup’s clients include BlackRock, KKR, and the U.S. Air Force.
The Issue Lies Beyond the Models
Sivulka assessed that AI agents often fail not due to model weaknesses but because of unclear instructions within companies. He claims that “about one in 100 employees” knows how to provide AI with enough context for quality work.
Poor instructions lead to “loops”: the agent tries to correct its actions, repeatedly accesses the model, and consumes additional tokens without proportional results. Sivulka described this as “spending tokens on spending tokens.” Ultimately, he concluded that companies are facing a management crisis rather than a technological one.
Fortune linked this thesis to a broader discussion on AI expenses. According to the publication, at the beginning of 2026, several companies tracked and encouraged token consumption as a measure of employee activity.
Later, some of these initiatives were rolled back. Previously, the publication reported that Meta and Amazon used internal token consumption ratings but abandoned them after employees began using AI for metrics rather than results.
Expenses Have Become a Separate Risk
Fortune cites UBS Global Research data indicating that nearly all AI company executives at a closed bank event discussed the issue of token expenses in a corporate setting.
One unnamed firm told UBS that its expenses on Anthropic rose from $20,000 in December to nearly $1 million in July. Management did not fully restrict usage but began implementing internal limits and warnings about threshold breaches.
According to journalists, UBS previously assessed concerns about token expenses as a significant issue for about 60% of organizations. The publication also mentioned an unnamed company that, according to Axios, received a bill of approximately $500 million in one month for using Claude after launching AI tools without limits.
Companies Shift to Model Management
Sivulka sees the solution not in abandoning AI but in developing management infrastructure for agent systems. This includes clear process descriptions, quality assessments, cost control, and task distribution among different models.
Fortune reports that some companies are already moving towards model routing. In this approach, simple operations are assigned to cheaper or faster models, while expensive advanced ones are used only for key scenarios.
Sivulka also warned of a new internal issue related to employees retaining important work context. He noted that people might be reluctant to transfer knowledge and processes to AI systems that make them indispensable. The expert emphasized that without trust and clear rules, workers will not assist systems that could change them.
In July, Financial Times columnist Sarah O’Connor described employee knowledge as a resource for corporate AI.
