Watch Your AI Spending
To curtail costs, a growing number of firms are imposing strict limits on employees' use of AI. Could this backfire?
You have a set amount to spend on AI this month—say, up to four lattes’ worth. Use it wisely.
Wary of the escalating costs of their AI investments, an increasing number of organizations are imposing limits on their employees’ use of AI. In some cases, firms are reducing the allowable use (and expense) of the tech to a fraction of its normal level. The move could make sense in the short term, experts say, but it could also cause multiple headaches down the road. “If I can’t innovate with it, or it runs the risk of stopping midwork, then I won’t rely on AI,” says Bryan Ackermann, Korn Ferry’s head of AI strategy and transformation.
In some organizations, hitting the new limit triggers a conversation in which employees must justify exceeding the limit to their managers (or the team tasked with implementing AI within the organization). In other companies, it’s a hard cap, with an employee cut off from AI use for the rest of the month.
Experts say it’s not surprising that companies have become more cost conscious around their AI spending. US firms are forecast to spend $280 billion on AI this year alone, over 50% more than what they spent in 2025 (these numbers don’t include the spending of the AI model-makers themselves). Each time an AI answers a question, develops a financial model, or anything else, it spends “tokens.” This spring, many AI creators moved away from a flat-fee pricing model to a per-token one. The cost of some software subscriptions instantly jumped by between 20% and 40%. Other subscription plans, which once allowed for unlimited AI use, now have hard limits, forcing firms to buy more tokens, much as someone might have to buy more minutes for a cell-phone plan.
Over the summer, many organizations placed caps on their technology employees’ use of AI, since software developers were racking up the biggest bills. But the latest round of limits extends to employees outside the IT department. Some organizations have found that workers are using pricey AI tools to perform low-value tasks, such as converting PDFs into slide presentations. In those so-called “token-chewing” cases, the employee might save time and effort, but the expensive end product doesn’t help clients.
Ackermann says the neither the limits nor the conversations about exceptions they’ve necessitated are practical for firms. For instance, if 2% of the workers in a 20,000-person organization need to discuss the rationale for their AI use, that would entail 200 additional meetings each month. “It’s unsustainable,” Ackermann says. The system also puts the burden on employees to keep track of their usage, which many have neither the tools nor the knowledge to do.
The new limits also could curtail people from experimenting with AI. Until recently, training was largely about letting employees develop their own ways of effectively using AI; now it may focus solely on teaching them how to deploy AI exclusively in ways that have already been shown to increase productivity. “You’ve had three years (of AI of throwing things at the wall and see what sticks,” says Janet Mertens, managing director of AMS Research at the Korn Ferry Institute.
Ackermann, who speaks with the AI model makers regularly, says they understand that pricing is an issue. The onus is on the developers to make AI for their corporate customers that’s affordably and dependably priced. “Otherwise, it will only be used as a toy,” he says.
Learn more about Korn Ferry’s AI in the Workplace capabilities.
