U.S. Census survey shows first decline in AI adoption since 2023

A recent U.S. Census Bureau survey shows that AI adoption has dropped for the first time since tracking began in November 2023. Large companies, those with more than 250 employees, used AI less often, falling from 13.5 percent in June to 12 percent in August. This change breaks the steady growth seen since late 2023. The survey includes data from 1.2 million U.S. firms.

The survey, called the Business Trends and Outlooks Survey (BTOS), is conducted biweekly by the Bureau. While the overall corporate AI usage for goods and services production still grew from 3.9 percent in late 2023 to over 5 percent by mid-2024, the adoption rate among mid-sized companies (firms with 20-250 employees) remained flat or declined. In contrast, only very small businesses (with fewer than 4 employees) showed a slight increase in usage. It is important to note that, since the surveys are conducted biweekly, these results could reflect a short-term blip rather than a lasting trend.

This slowdown points to some of AI’s limits. Most companies surveyed, about 95 percent, said they have not made new revenue from using AI. Recent studies also show that AI adoption has led to a 13 percent drop in jobs for young workers, with junior positions affected more than senior ones. Companies still need people to oversee AI because it can make mistakes, which means it is not always a way to cut costs. On top of that, OpenAI’s new GPT-5 model did not do well in benchmark tests, showing more challenges for the technology.

Despite these issues, tech stocks continue to hit record highs, even as AI usage declines. Nvidia's share prices, for example, dipped after an earnings call about sold-out AI GPUs. Meanwhile, analysts are wary of a potential “AI winter” that could dampen the hype surrounding the industry. Reflecting shifting sentiment, some companies are even rehiring workers after realizing that AI isn’t as good as they initially thought.

Looking forward, it is unclear if current levels of AI investment can last, especially since most enterprise AI projects are still not making a profit. More long-term data will be needed to see if this recent dip is just a short-term change or the beginning of what some analysts call an "AI winter."