"We discovered that an employee ran a procedure at night that cost 38 thousand dollars"

Until not long ago, organizations knew how to calculate almost exactly how much human resources would cost them each month. The number of employees, salaries, and positions were the basis upon which budgets and work plans were built. However, with the accelerated entry of artificial intelligence tools into organizations, a new cost unit was added to the equation: tokens. A token is the unit of measurement by which AI models calculate the extent of their use. Every question sent to the model, every document it reads, and every answer it returns are translated into tokens, based on which the organization is charged.

The more advanced the model or the more complex the task, the higher the token consumption and cost. While in the past this was a technological term known mainly to engineers, today it is becoming a concept that every manager must understand. More and more AI services are offered in a usage-based pricing model, making AI expenses a significant budget component. Managers are now required to learn to manage not only human resources but also the digital "workforce" that AI models produce.

"Tokens are the smallest unit of consumption that AI mechanisms know how to work with. They measure data traffic, whether it is a question I ask or an answer I receive," explains Inbal Namir, Managing Director at Deloitte and leader of the Future of Work practice. According to her, organizational economics is now composed of two factors: human and digital. While the former is known and certain, the latter requires new capabilities to predict costs as accurately as possible.

This change also affects how organizations evaluate efficiency. If in the past they measured output primarily in relation to working hours, today a new variable has been added: the amount of AI resources consumed. Employees who are not experienced enough or lack awareness of the fact that tokens consume money may work inefficiently. Consequently, organizations are starting to add evaluation mechanisms that measure output against the number of tokens consumed.

Oded Tahori, CEO of Gene Technologies, notes: "An organization discovers that an employee ran a procedure at night in three hours that cost 38 thousand dollars. This is something unplanned, unbudgeted, and certainly not approved. When there is no system that blocks this, and there are thousands of employees, it can reach huge sums of millions."

As the use of artificial intelligence expands, organizations can no longer settle for generally encouraging employees to use these tools. They must set a clear budget policy, decide who is allowed to exceed it, and ensure that investments in AI resources indeed create business value. Ultimately, as Dor Atias, founder and CPO at Cycode, points out, the final product remains the most important metric. Tokens have become part of a new organizational language that affects budgets, work processes, employee training, and management decisions.