As employees use artificial intelligence (AI) for simple tasks and budgets are quickly depleted, companies are moving to control costs.

TechCrunch reported on Tuesday that Accenture has begun steps to prevent employees from exhausting internal token budgets by overusing AI for basic work.

The key point is that the focus is shifting from expanding AI use to managing cost efficiency. Until early this year, the AI industry encouraged companies to adopt AI aggressively up to their budget limits. Some companies even ran leaderboards to pit internal AI usage against one another, but once operations began, awareness has grown that costs are significant and results are not clearly visible.

Within Accenture, employees are reported to have increased token use by using AI for relatively simple tasks, such as converting PDFs into presentation slides. The adjustment is drawing more attention because it follows Accenture's earlier warning to employees that they could face disadvantages in promotions if they do not use AI.

Management attention is also moving from expanding use to verifying returns on investment. Justice Kwark (저스티스 콰크), Accenture's head of agentic AI strategy, explained that executives at the chief technology officer (CTO), chief operating officer (COO) and chief information officer (CIO) level keep asking whether they are getting real value from the money spent on AI. For companies, the more important management task has become how much AI to use for which work, rather than AI adoption itself.

This trend also dovetails with recent moves to cut AI-related costs. Cases are continuing to emerge of companies readjusting AI budgets, and the burden of token costs is spreading into broader doubts about AI business models. The 'AI sell-off' in recent days that has shaken the share prices of some AI-dependent companies and memory chipmakers was also cited as reflecting such concerns.

Ultimately, the AI industry has reached a testing ground after the expansion phase. It has become harder to maintain budgets on expectations alone that it is a new and interesting technology, and companies have entered a phase in which they must justify costs with actual results. The Accenture case shows that, separate from increasing employee usage, management standards are needed on which tasks should use AI to deliver cost-effective results.