US companies have accelerated their adoption of generative artificial intelligence (AI) over the past two years, but actual results have varied significantly across organizations.

According to multiple recent surveys of executives, many companies have struggled to translate AI investment costs into tangible business results.

Meanwhile, some companies have achieved visible outcomes by adopting AI focused on clearly defined business tasks and establishing performance metrics that can measure return on investment (ROI).

▲ AI adoption spreading rapidly…over half of companies investing more than $1 million annually

According to the International Business Times (IBT) on the 26th (local time), Writer, a generative AI company, conducted a survey with independent research firm Workplace Intelligence targeting 1,200 non-technical employees and 1,200 C-level executives who use AI in their work.

According to the survey results, 59% of responding companies reported investing more than $1 million annually in AI. AI utilization was also expanding rapidly.

97% of responding executives said they had adopted AI agents over the past year, and 52% of employees were already using them in their work. Additionally, 94% of C-level executives and 70% of employees were using AI tools for at least 30 minutes daily, while 64% of executives reported using AI for more than 2 hours a day.

▲ AI utilization increasing but investment returns falling short of expectations

Although AI usage increased rapidly, it has not translated into the business results companies expected.

75% of C-level executives evaluated their company's AI strategy as "more for show than as substantive operational guidance." 48% of executives said they were disappointed with the results of AI adoption, while 39% of companies had not even developed a formal strategy for monetizing AI investments.

In fact, only 29% of companies reported achieving meaningful returns on investment through generative AI.

▲ Security incidents and employee resistance also create serious side effects

The survey confirmed various negative side effects accompanying AI expansion.

67% of executives answered that data breaches or security incidents may have already occurred due to unauthorized AI tool usage. 35% of employees admitted to having entered confidential company information into public AI services.

Additionally, 55% of respondents characterized their company's AI usage environment as an "uncontrolled chaotic state." 36% of companies lacked formal frameworks for managing and supervising AI agents, and 35% were unable to establish response systems capable of immediately halting malfunctioning AI agents.

AI adoption was also increasing anxiety within organizations. 73% of CEOs reported experiencing stress or anxiety related to AI, and 64% expressed concern about losing their positions due to AI transformation failure.

69% of companies were planning workforce reductions through AI, yet 39% had not developed strategies to generate new revenue through AI. 29% of employees, and particularly 44% of Generation Z, reported having intentionally hindered their company's AI strategy.

▲ Evaluation focused on 'adoption performance' exacerbating negative effects

The survey results suggest that many companies have evaluated AI adoption itself as an achievement under pressure from investors, boards, and industry-wide AI expansion.

In fact, some companies adopted the number of AI licenses issued or employee AI usage volume as key performance indicators (KPIs). However, this approach was analyzed as having led to unforeseen side effects including increased unauthorized AI tool usage, security incidents, and employee resistance.

▲ Enterprise AI market continues growth…demand concentrating on practical solutions

However, there were also analyses suggesting that corporate AI adoption is not slowing down.

According to a survey by venture capital firm Andreessen Horowitz (a16z), 29% of Fortune 500 companies and 19% of Global 2000 companies had already converted into paying customers of major AI startups.

This result was interpreted as showing that companies are concentrating investments on commercial AI software applicable to actual work rather than simple pilot projects.

Artificial Intelligence [Reuters/Yonhapnews provided]

▲ Common traits of high-performing companies: 'clear tasks' and 'measurable goals'

After analyzing revenue data from AI startups, a16z found that successful enterprise AI adoption tends to concentrate on specific tasks and industries where performance measurement is easy.

The most representative application area was software coding, followed by customer support and enterprise search services. By industry, IT, legal, and healthcare sectors showed the most active adoption of AI.

These tasks generally share common characteristics of being text-focused, highly repetitive, and having easily verifiable results, making them well-suited for measuring AI utilization effectiveness.

▲ Phased approach over company-wide transformation yields higher investment returns

According to survey results, companies with successful AI adoption chose a strategy of starting projects centered on clearly defined business processes and establishing performance measurement metrics, rather than pursuing company-wide transformation from the outset.

In contrast, companies that pursued company-wide AI transformation before establishing AI governance, work scope, and success criteria frequently reported lower investment returns and increased operational confusion.

▲ Facing limits of 'show-only' adoption amid expanding AI investments

As AI investment scale continues to increase, the costs of "show-only AI strategy" that merely highlights the fact of AI adoption are expected to grow.

Experts analyzed that companies that rushed into adoption without first identifying areas where AI can actually generate measurable business value would face a situation where they must answer fundamental questions they had avoided from the beginning of the AI innovation period—namely, "where can AI actually generate return on investment?"