Bosses Replace Staff with AI to Cut Costs, But End Up Spending More
Jakarta, CNBC Indonesia - Many companies were initially enthusiastic about replacing or reducing human labour with artificial intelligence (AI), hoping to cut long-term operational costs. But the bitter reality has now begun to emerge: AI usage bills have swelled, often exceeding the wages of the employees who were laid off.
A recent report from the leading accountancy firm KPMG, released in early July 2026 and cited by the technology sites Futurism and The Register, has opened the eyes of corporate leaders to the fact that AI is not a cheap solution that can be used for free.
During trial periods, AI service providers usually offer fixed-price contracts or steep discounts to entice companies to give it a try. However, as computing power costs have surged, nearly all large language model developers have now shifted to pay-as-you-go systems, or billing calculated per unit of data processed.
This means the more often AI is used, the larger the bill. Many board members who initially assumed the costs were fixed are now shocked to receive invoices that keep rising sharply.
KPMG’s survey, involving 2,145 senior executives from 20 countries, uncovered some startling facts. As many as 29% admitted they did not know exactly where the increase in AI costs came from, while 33% said their lack of understanding of AI economics was itself becoming an obstacle. Only around a quarter of companies have clear and measurable systems for overseeing AI budgets.
“Many organisations are still building the capability to forecast, monitor and manage AI spending accurately,” the KPMG research team wrote in its report.
The CEOs’ initial logic was simple: replace employees with AI so that the costs of salaries, benefits and leave would disappear.
Reality, however, has proved different. AI requires server infrastructure and electricity that do not come cheap, specialist expertise, and still requires human intervention to “fix” AI errors.
As a result, AI usage bills - measured in tokens - have ballooned severalfold because usage went unbounded. In some reported cases, companies ended up paying for AI at a value of two to three times the payroll budget of the teams they had previously dismissed.
Joe Wilkins, author of the report at Futurism, noted that AI has more often been used as an excuse to weaken employees’ bargaining positions rather than as a genuine efficiency solution.
“Many bosses treat AI like a tool that can be plugged in and will run without ongoing costs. In fact, AI requires a continuous stream of computing power costs, just like electricity or water,” he wrote.
As a consequence, cost-cutting strategies that rely on AI are now being reconsidered. Many companies are beginning to realise that retaining a trained workforce is actually more predictable in cost than depending on a system whose bills are unpredictable.