(Yicai) July 21 -- Chinese companies have accelerated the deployment of artificial intelligence over the past year, but many are finding that rising investment has yet to translate into proportional gains in productivity. Executives from insurance giant Taikang Insurance Group, home appliance maker Haier Group and other industry leaders discussed this challenge at a forum hosted by Irish consulting firm Accenture held on the sidelines of WAIC.

"There is a lot of talk about AI, but in reality, the output is not that high," said Du Yanbin, a member of Taikang Insurance’s management committee and general manager of its technology management department.

“What we are seeing is an efficiency paradox,” Du said. “When a new productivity tool first emerges, expectations for its value are very high. However, once it is deployed at scale, the efficiency gains are often much smaller than anticipated.”

Accenture China teamed up with Haier’s smart home solutions unit Haier Smart Home to calculate the length of the production cycle at one factory, from receiving an order to shipping the finished product, a process requires 12 stages of approval, said Jason Cheung, managing director of Accenture China’s supply chain and operations division.

"The entire cycle took 7.1 days, but the actual time required to turn raw materials into finished products was only 2.3 days," Cheung said. “The remaining time was spent moving through layers of approvals.”

Haier Smart Home is restructuring its organization to streamline these processes, said Sun Danfeng, vice president of the Qingdao-based firm. “If the organization does not change, the individuals responsible for the original process nodes will no longer have meaningful work to do, leading to friction.”

One reason why many AI projects have failed to meet expectations is that companies often begin with overly ambitious goals instead of starting with narrowly defined, manageable applications and expanding gradually, Du said.

He gave the example of Taikang Insurance’s AI tool for financial approval approvals, which went live in June. Only after the system had operated reliably for a month and achieved an accuracy rate of more than 99 percent did the Beijing-based company begin the process of scaling the tool across similar approval tasks.

Taikang Insurance also developed its own AI sales training simulator that uses voice interaction to help insurance agents repeatedly practice their customer communication skills. After about 10 training sessions, participants were able to improve their test scores from the 20-point range to a stable score of around 70.

AI-guided drones are now being used for field inspections at PepsiCo's farms in southern China, said Li Youlin, senior vice president of supply chain for the US beverage giant’s Asia Pacific and China divisions. Previously, one person could only inspect 3.3 hectares of farmland, but now the process is fully automated, delivering exponential reductions in costs.

Empowering Employees

Internal company data shows that employees who receive promotions use AI about 50 percent more frequently than other staff, said Victoria Wu, vice president of human resources for US tech firm Cisco’s China arm. Employees identified as “AI experts” are also more than 40 percent more likely to be classified as key talent.

Otis has adjusted its recruitment process, said Kathy Kang, senior vice president of human resources for the US elevator manufacturer’s China division. Instead of determining the headcount first and recruiting afterward, the Connecticut-based firm now starts with customer demand and identifies the skills required before deciding on staffing needs. Elevator maintenance technicians can now identify spare parts by taking photos and instantly retrieve historical maintenance records.

Accenture has built skills profiles for over 700,000 employees worldwide, with an accuracy rate exceeding 93 percent, said Catherine Huang, president of human resources for Accenture China. The Dublin-based company has also created a new "Talent Futurist" role to anticipate the capabilities needed by the organization in the future.

"We should not make employees the objects of AI-driven transformation, but rather make them collaborators in this transformation," she said.

Regional Spillover

The organizational changes driven by AI are also spreading beyond China.

PepsiCo is introducing predictive maintenance at new factories in Vietnam and Indonesia, reducing the number of maintenance staff to about 20 from over 30, Li said. In a typical factory with around 500 employees, roughly 8 percent of the workforce is dedicated to specialized maintenance. However, as smart sensors and automated work-order systems are deployed, these processes are becoming increasingly streamlined.

A Thai bank with approximately CNY700 billion (USD103.5 billion) in assets decided to launch a comprehensive AI transformation after its board of directors visited Shanghai last year, said Yu Yi, Accenture’s global vice president and president of its digital core division for Southeast Asia. The lender organized hackathons involving 700 to 800 employees. Initially, the focus was on basic prompting, however, two weeks ago, the initiative was upgraded to include agent applications.

Editor: Kim Taylor