(Yicai) Sept. 23 -- Artificial intelligence is accelerating its entry into production applications in China, but almost half of companies say actual returns after deployment remain below expectations, with many noting they cannot clearly calculate the return on investment from AI projects, according to the findings of a new survey.

The survey, which included 205 executives from various industries in China, most of whom were strategic decision-makers, found that AI deployment has failed to meet return expectations for 49 percent of respondents, according to a report released by Deloitte China and the HKU Center for AI, Management and Organization yesterday. Over 44 percent said that due to the scope of AI applications continuing to expand, their ability to quantify value and their evaluation frameworks have not kept pace.

In addition, the percentage of companies in the AI exploration and pilot phase fell to 25 percent this year from 37 percent last year, while that of those in the production deployment phase rose to 64 percent from 56 percent, the report showed. Companies' expected return on investment is mainly in the 10 percent to 25 percent range, while their actual returns are more often below 10 percent.

While AI adoption and production deployment among firms continue to increase, widespread use of AI tools does not mean companies have completed their transformation, said Pascal Hua, national managing partner of Deloitte consulting businesses in China. Many companies have introduced AI but continue to rely on existing business processes and operating models, so the tech often stays at the efficiency-enhancement level, he pointed out.

In the next phase, the focus of competition will shift from technology adoption to value realization, Hua noted, adding that the factor that will truly differentiate companies in the future will be their ability to integrate AI into core business processes, moving applications from mere tool additions to workflow transformation, and consistently generating measurable and verifiable business outcomes.

Behind the disappointing returns is the fact that most companies lack effective mechanisms to assess the benefits of AI applications, with 60 percent still lacking a systematic framework for value measurement, the report said. Without a unified measurement standard, firms struggle to determine which projects are worth replicating and which to scale back. In response, they can optimize AI project management through three key areas: value assessment, phased evaluations, and investment management.

AI applications will further reshape firms' human resource composition. The survey found that 65 percent of respondents expect fewer entry-level or junior roles, while mid-level positions will likely face structural reorganization rather than a simple decline, with greater pressure amid their functions shifting from supervision to coordination and judgment. Top-level posts should remain stable because strategic decision-making and accountability are difficult to replace, the respondents pointed out.

Editor: Martin Kadiev