(Yicai) July 17 -- While the percentage of Chinese companies fully adopting artificial intelligence nearly doubled, the percentage of them achieving a significant boost in financial performance remains very low, mainly because of the impact of the so-called Token Tax, according to a report by global consulting giant Accenture.

Eighty-eight percent of Chinese firms that carried out AI pilots last year have actually moved past trials, compared with 46 percent the year before, according to the China Enterprise Digital Transformation Index report released by Accenture yesterday. Meanwhile, the ratio of companies that managed to translate AI investment into revenue, profit, or efficiency improvements increased only to 14 percent from 9 percent.

The China Enterprise Digital Transformation Index report covered 160 Chinese enterprises across eight major industries, including high-tech manufacturing, pharmaceuticals, and automotive. It was released ahead of the World AI Conference, which is being held in Shanghai from today to June 20.

The gap between the scaling of AI applications and the speed of realizing value was mainly created because of the increase in prominent hidden costs, especially the Token Tax. In fact, Accenture found that the token usage for the same exact task can vary up to 17 times due to differences in breakdown methods, tool-calling paths, and model selection.

After adopting AI, Chinese companies' annual budgets increase by tens of millions of Chinese yuan, equal to millions of US dollars, without visible returns, said Cao Qifeng, president of Accenture's AI and data practice in China. The problem does not lie in the models' capabilities, but in whether the firm has figured out what problem AI is meant to solve and how to measure the results, he explained.

"Applying AI internally is not merely a tooling issue, but it requires top-level strategic planning," Cao noted.

This issue is not unique to Chinese companies. For example, Uber reportedly exhausted its entire annual AI budget within four months and had to set a cap on employee token usage, while Meta and Amazon canceled their internal token consumption rankings.

Firms’ enthusiasm over the past year for ranking employees or teams by token consumption is like putting the cart before the horse, said Qiu Jing, head of Accenture Research in China. However, they have started to realize the need to deploy different models for different tasks, based on different business scenarios and task complexities, he noted, adding that this conveys a return to a results-oriented approach.

Only 10% of Firms Have Capabilities for Next-Stage Competition

China's digital transformation index has reached 59 points this year, expanding at the fastest pace in nine years, according to the report. Against this backdrop, Accenture created an assessment framework based on three future capabilities -- autonomous intelligence, value density, and multi-polar operations -- to measure whether enterprises possess the capabilities for the next stage of competition.

The results show that only 10 percent of surveyed enterprises had a digital transformation index in the top 50 percent of their industry, while having at least two of their three future capabilities in the top 20 percent of their industry.

The proportion of enterprises where AI agendas are led by their top management has risen to 61 percent this year, yet implementation remains superficial. In fact, of the 63 percent that have deployed AI assistants, only 18 percent have advanced systematic process redesigns.

AI is not purely an information technology issue or a business  department issue, it requires cross-department and cross-functional collaboration to be truly implemented, Cao said.

He compared this transformation to the shift from steam engines electric motors in the 19th century. Factories initially simply stuffed electric motors into the original steam engine rooms without changing the power transmission methods, which did not improve production efficiency. Instead, efficiency only improved once the entire set of processes was redesigned around the new form of power.

From Mere Tool to Growth

Eighty-two percent of Chinese enterprises hope to use AI to open new spaces for business growth, not just to reduce costs, Accenture's report showed. However, the proportion of enterprises truly achieving scaled value -- a production efficiency increase of over 10 percent or a revenue or profit increase of over 5 percent -- remains at only 14 percent.

How to transform AI from a tool into a growth driver can be seen in a few case studies.

For example, Taikang Life Insurance structured underwriting rules for over 1,000 products, a 1.8-million-entry medical knowledge graph, and expert reasoning experiences from 50,000 historical underwriting cases, integrating them into an AI system. As a result, the underwriting accuracy rate increased to over 95 percent, allowing agents to obtain complete underwriting advice within one minute of uploading a medical examination report.

"After AI emerged as a new productive force, the original organizational structures, processes, and skills have not yet changed, which is a very big conflict and a core obstacle to scaling AI," said Du Yanbin, member of Taikang Life Insurance's management committee.

GE Healthcare China began building a platform for data localization and enterprise-level data six years ago. Under a globally unified governance standard framework, the Chinese team independently advanced local AI governance, which is now being reverse-exported to GE Healthcare's global system.

Faced with a saturated charging network and intensifying hardware homogenization, charging pile operator Star Charge shifted away from relying on the number of piles to drive growth to using AI to manage the highly dynamic, multi-variable dispatching between photovoltaics, energy storage, and charging.

For instance, an industrial park connected to the AI charging system allows users to save 21 percent in charging costs, while increasing the station's operating revenue by 22 percent, according to Star Charge .

AI as New Gateway for Global Expansion

More and more Chinese companies are foraying into overseas markets to ease the pressure from domestic competition.

Fifty-nine percent of the firms surveyed by Accenture this year believe local  regulatory and compliance requirements have significantly increased, and 58 percent said that competitive pressure from domestic peers has risen. The two figures increased from 33 percent and 26 percent, respectively, last year.

China's outbound direct investment rose 7.1 percent to USD174.4 billion last year from 2024, with over 50,000 overseas enterprises reaching 190 countries and regions, according to data from the Ministry of Commerce.

AI agents are providing a new gateway  for Chinese brands, said Qiu. As overseas consumers' decisions are increasingly completed via comparisons and recommendations by agents, the first-mover advantage accumulated by legacy brands is being partially leveled. Product strength and service quality are now more directly recognized and recommended by intelligent agents.

This is crucial for Chinese enterprises seeking to venture abroad. Accenture's survey shows that even though over 60 percent of Chinese enterprises aim to become global leaders, only 5 percent possess global brand influence, and 12 percent can influence the development direction of international industries.

Behind this gap, there is a difference in narrative styles. "In other countries, you need to have an altruistic narrative, demonstrate how you are helping the local ecosystem, enterprises, industries, and consumers achieve better growth or value, rather than just exporting products, technologies, and standards," Qiu explained.

The next stage of the search for value through AI is not continuing to optimize along the existing growth curve but using AI to reconstruct capabilities and value-creation logic, expanding new growth spaces in a multi-polar global market.

Leading in the AI era is not a technological but a leadership choice, said Samantha Zhu, global vice president of Accenture and chairman of Accenture China. Enterprises need to seamlessly integrate technological capabilities, business scenarios, organizational mechanisms, and governance systems so that AI can transform from a single-point efficiency tool into a strategic engine driving high-quality growth, she added.

Editor: Futura Costaglione