Three years after generative AI ushered in a "new era," a huge gap has emerged between market hype and actual corporate implementation. A latest AI application survey report released by McKinsey in December reveals a harsh reality: although everyone is talking about AI, almost no one is making significant profits from it.

Adoption does not equal profitability: 88% of the "false prosperity"

Data shows that 88% of companies have integrated AI into at least one function on a regular basis, a proportion that has further increased from 78% last year. However, the depth is seriously lacking — nearly two-thirds of respondents admitted that their companies have not yet fully launched large-scale AI deployment, with most institutions still in the exploration or pilot phase.

The more critical issue is that only 39% of respondents believe AI has had a substantive impact on the company's EBIT (Earnings Before Interest, Taxes, Depreciation, and Amortization), and most of these contributions are less than 5%. This means that over 60% of companies are currently "spending money for publicity."

McKinsey warns investors: don't be deceived by the "AI content" in corporate earnings calls. The current AI applications show a typical "broad but shallow" characteristic — capital expenditures are happening, but return on investment is severely lagging.

Artificial Intelligence, Robotic Arm, AI (1)

What did the 6% winners do right?

The survey defines about 6% of companies as "high-performing AI companies" (EBIT increased by more than 5% due to AI). These companies differ fundamentally from the other 94%:

  • Rejecting pure cost-cutting: 80% of ordinary companies focus only on efficiency, while high-performing companies pursue growth or innovation simultaneously
  • Reengineering workflows: They are nearly three times more likely to fundamentally reshape workflows
  • Substantial investment: One-third of them invest more than 20% of their digital budget in AI, which is nearly five times that of other companies
  • Executive involvement: Establish clear human-machine collaboration processes and manual verification mechanisms

AI agent hype cools down, China stands out

The hotly touted "AI agents" are still in early experimental stages. 62% of respondents are testing them, but only 23% have started large-scale implementation in at least one function. Notably, 45% of enterprises in mainland China have achieved large-scale AI implementation or full deployment, higher than the global average of 38%, and 83% of enterprises have routine use of generative AI, showing stronger execution ability.

Employment and risks: Layoff expectations rise

32% of respondents expect employee numbers to decrease by more than 3% in the next year, while only 13% expect growth. 51% of application companies have encountered at least one AI negative incident, with the most common being "inaccurate results" (30%), i.e., AI hallucination issues.

McKinsey advises investors to be cautious of companies that only remain at the "pilot" stage, and to look for "real players" who are willing to restructure business processes and make substantial investments. The AI narrative in 2025 must shift from "who is using" to "who is profiting."