AI Trends

Nankai Scholar Points Out That Over-Reliance on AI for Human Substitution Leads to Service Degradation and Employment Risks

YicaiUpdated Nov 18 20:13

AI summary

Scholars from Nankai University have published an article pointing out that some companies are over-applying AI customer service and review systems with the goal of reducing costs, leading to a decline in service quality, harm to consumer rights, and structural employment risks. The article recommends establishing a pre-assessment system for employment impacts, setting rigid human fallbacks in key scenarios, and improving mechanisms for job transition training and unemployment protection.

Why read this

AI score 64

The article analyzes the hidden dangers of AI replacing human workers through policy and case studies, proposing the establishment of an employment-friendly evaluation mechanism and clear boundaries for human-machine collaboration.

Topics

Tags

Body · AI translation

Artificial intelligence (AI) has rapidly moved from assisting in content generation and handling standardized tasks to participating in production and operation links such as customer service, financial auditing, logistics scheduling, and labor management. While intelligent transformation should theoretically promote enterprise efficiency improvement and better working conditions, some enterprises measure the success of technological upgrades by reducing labor input, leading to phenomena such as ineffective performance of service systems, premature compression of manual positions, and inadequate personnel placement following technology deployment. In October 2026, the General Office of the State Council issued the "Opinions on Adapting to the Development of Artificial Intelligence Technology to Assist in Building an Employment-Friendly Development Model," explicitly requiring effective mitigation of the employment substitution impact of AI applications. On November 3, Wang Ruomeng, Deputy Director of the Department of Innovation and High-Tech Development at the National Development and Reform Commission, further proposed closely monitoring the impact of AI on employment structures, strengthening early warning mechanisms for employment substitution risks, and ensuring bottom-up, inclusive social security. Currently, in practice, some enterprises' intelligent transformations exhibit a tendency to prioritize cost reduction over efficiency enhancement. The cost savings within enterprises are not synchronized with improvements in overall social benefits, and related risks are transmitting to areas such as consumer rights, labor relations, and social employment.

Currently Prominent Issues

(1) Excessive substitution of human labor by intelligent services coexists with declining service quality alongside corporate cost reduction. AI customer service is accelerating its penetration into sectors such as e-commerce, telecommunications, and finance. Some enterprises focus primarily on compressing labor costs, neglecting the actual service capabilities of intelligent systems. Data from the China Consumers Association shows that in the first half of 2026, national consumer association organizations received a total of 986,000 consumer complaints, with after-sales service issues accounting for the highest proportion at 26.79%. New hotspots for complaints include false promises by AI customer service, difficulties in accessing human agents, and inaccurate generated content. Some enterprises even use AI interception rates rather than problem resolution rates as assessment indicators. Swedish fintech company Klarna serves as a typical negative example: in 2024, it prominently launched an OpenAI-powered customer service system, handling 2.3 million conversations in its first month—equivalent to 700 full-time customer service representatives—and reduced response times from 11 minutes to 2 minutes, expecting annual savings of approximately $40 million while significantly reducing its workforce. However, by 2025, its CEO publicly admitted that "excessively high cost proportions led to a decline in service quality." Unable to handle complex disputes, emotional soothing, and high-risk scenarios, the company was forced to restart hiring for human customer service staff, retreating from "full substitution" back to a "human-machine layered" model. When intelligent systems cannot understand complex demands and human entry points are hidden behind layers, the reception costs saved by enterprises essentially transform into burdens for consumers involving waiting, repeated explanations, and recurrent complaints. Intelligent customer service should handle standardized affairs rather than becoming a barrier blocking necessary human services. Transformation driven by short-term report dividends saves current labor costs but overdrafts service resilience, customer trust, and organizational capability, thereby damaging long-term corporate competitiveness.

(2) Intelligent transformation tends to focus on job compression, with a disconnect between technological substitution and actual efficiency gains. AI is extending into areas such as professional review and quality management. Some enterprises view reducing headcount as a technical upgrade, ignoring the labor value of tasks that cannot be easily quantified, such as handling exceptions, reviewing results, and exercising experienced judgment. In April 2026, the Hangzhou Intermediate People's Court released typical cases regarding the protection of rights and interests for AI companies and their employees. A fintech enterprise demoted and reduced the salary of Zhou Mou on the grounds of AI technology upgrades and the claim that original quality inspection work could be completed by AI. After Zhou Mou refused, the enterprise notified him of the termination of his labor contract. The court ultimately ruled that the enterprise had illegally terminated the contract and ordered it to pay compensation of over 260,000 yuan. The fact that technology can complete certain work tasks does not mean it can completely replace original positions. Especially in highly specialized links such as quality control, manual review not only serves an error-correction function but is also an important link in ensuring business reliability and controlling operational risks. If enterprises merely reduce personnel based on technological substitution capabilities without evaluating system error rates, the necessity of manual review, and comprehensive operating costs, they may create a situation where short-term labor cost decreases lead to long-term increases in operational risk. Overly pursuing staff reduction and cost cutting may weaken quality supervision and error-correction capabilities, causing technological applications to deviate from the goal of actual efficiency gains.

(1) The short-term cost reduction for enterprises conflicts with long-term operational benefits, and the risk of low-quality intelligent transformation is rising. Enterprises have the right to use new technologies to improve operational efficiency; however, when calculating returns, they may only account for reduced labor costs while ignoring subsequent expenses such as system maintenance, error handling, customer churn, and labor disputes. In October 2026, media outlets conducted practical tests on the customer service hotlines of 16 banks, finding that some banks' intelligent voice menus were nested in layers, making it difficult for customers to access human agents even after multiple operations. Banks view intelligent customer service as a key tool for reducing costs and increasing efficiency, but the difficulty in accessing human agents has instead become a new issue affecting customer experience. The labor costs saved internally by enterprises may be offset by customer churn, complaint handling, and service recovery costs. If companies continue to replace process optimization and talent development with workforce compression over the long term, their professional service capabilities and sustained competitiveness will be weakened.

(2) Weakening of enterprise service responsibilities leads to increased costs for protecting consumer rights. Intelligent customer service not only handles business inquiries but also controls consumers' channels to access human services. Once enterprises excessively pursue diverting calls from human agents, technical systems may become obstacles to after-sales rights protection. According to case analyses published by the China Consumers Association (CCA) regarding complaints received by national consumer organizations in the first half of 2026, an AI customer service agent on a certain car trading platform promised that the vehicle price included transfer fees, yet during the actual transaction, customers were required to pay an additional 1,000 yuan. While enterprises provide services through intelligent systems, they may shirk responsibility citing reasons such as algorithm errors or inconsistencies between human and intelligent customer service information, forcing consumers to bear more time and rights-protection costs. Consequently, AI applications not only affect service efficiency but may also cause distortion of transaction information, difficulties in liability determination, and blocked after-sales relief. By compressing service costs through intelligence, enterprises shift extra verification, communication, and rights-protection costs onto consumers, and may even use technical processes to reduce consumers' willingness to pursue accountability.

(3) The separation of job replacement gains from employment transition costs transmits structural employment risks to society. AI creates new jobs in engineering development, data governance, and equipment operation and maintenance; however, job creation and job replacement do not necessarily occur at the same time, in the same region, or for the same groups. Data from Zhaopin showed that in the first three weeks after the Spring Festival in March 2026, the number of positions related to AI agents increased by 455% year-on-year, while demand for some entry-level white-collar positions continued to shrink. Growth in emerging technology jobs does not mean that existing workers can simultaneously transition to these roles. Particularly in labor-intensive industries such as customer service, data processing, and logistics distribution, if enterprises centrally compress manual investment, workers will face skill conversion challenges, income interruption, and difficulties in re-employment, further transferring related costs to families and public employment security systems. The accumulation of short-term cost-reduction behaviors by individual enterprises may exacerbate structural employment contradictions and increase the overall adjustment costs for society.

Recommendations for fostering synergistic human-AI development

1. Establish an employment-friendly evaluation mechanism for AI applications and implement a pre-assessment system for key industries. Change the practice of evaluating project effectiveness based on model adoption rates, labor substitution rates, and single-item cost savings. Promote the simultaneous disclosure or internal recording by enterprises of metrics such as first-contact resolution rate, successful human-agent transfer rate, error correction rate, changes in customer complaints, net job changes, and personnel placement status. For large-scale intelligent transformations in labor-intensive industries such as customer service and logistics warehousing, establish a pre-assessment and periodic monitoring system for employment impacts, focusing on scenarios involving concentrated job exits, declines in service quality, and increases in labor disputes. For intelligent projects receiving government support, use actual efficiency gains and effectiveness in stabilizing and transitioning jobs as evaluation criteria, avoiding the use of labor substitution rates or cost savings to replace the true effects of technological upgrades. National policies have proposed employment impact assessment and risk monitoring mechanisms; implementation should focus on clearly defining the assessment subjects, data standards, trigger conditions, and rectification paths. Evaluations should not be guided by the notion that "the more labor is reduced, the more advanced it is," nor mechanically treat retaining more personnel as better. Instead, they should verify whether technology truly improves output quality and working conditions.

2. Implement scenario-based human-machine collaboration management, setting rigid human fallbacks for basic rights and high-risk links. Based on the service nature, technical risks, and degree of labor substitution in different industries, clearly define the boundaries of human-machine collaboration in categories. In fields such as e-commerce, telecommunications, and finance, retain convenient channels for manual handling of matters such as complaint rights protection, fund transactions, and fee disputes; smart responses must not replace necessary manual processing. In fields involving personal safety such as healthcare and elderly care, adhere to intelligent assistance with professional personnel leading; set requirements for manual review and emergency takeover for diagnostic decisions, emergency treatment, and care for special groups. In fields such as transportation and logistics distribution, retain capabilities for manual monitoring, intervention, and emergency response for abnormal situations in autonomous driving, intelligent scheduling, and equipment operation. In basic public service fields such as government affairs and public employment, ensure that the elderly, persons with disabilities, and groups with insufficient digital skills receive necessary offline manual services. Incorporate manual takeover capability, service quality, and abnormal handling effectiveness into application assessments and regulatory inspections for key scenarios. Research setting rigid human fallback requirements for critical links involving personal safety and major rights, preventing enterprises from excessively compressing manual investment under the guise of intelligent transformation, thereby shifting technical operational risks and service costs onto society.

(3) Strengthen enterprises' responsibilities for technological substitution of jobs, and improve mechanisms for buffering employment impacts and guaranteeing job transitions. First, establish a pre-placement mechanism for technological substitution. For enterprises implementing large-scale job adjustments due to the application of artificial intelligence, promote the simultaneous formulation of personnel placement plans before technology deployment, clearly defining affected positions, workforce scale, and transition arrangements. Prioritize stabilizing employment through internal transfers, job adjustments, and paid training, preventing arbitrary pay cuts and layoffs under the guise of technological upgrades, and strictly implementing statutory requirements for labor contract modifications, terminations, and economic compensation. Second, improve enterprise-led retraining mechanisms. The human resources and social security departments should collaborate with industry associations and key enterprises to identify new job demands created by AI. For positions significantly affected, such as customer service, data processing, and logistics warehousing, conduct skills training in related areas like intelligent equipment operation and maintenance and data quality auditing, ensuring training content aligns with actual corporate labor needs. Incorporate actual transfer rates, employment stability, and wage changes into the evaluation of training subsidies and steady-job support policies, preventing the use of training headcount as a substitute for actual placement effectiveness. Finally, improve the unemployment transition guarantee mechanism. For workers who must leave enterprises due to job displacement, promptly facilitate the application for unemployment insurance benefits, vocational training subsidies, job recommendations, and employment assistance coordination. Explore transitional arrangements that synergize enterprise transition support with public employment services, implement targeted assistance for groups facing employment difficulties, reduce income loss during the period from job exit to re-employment, and prevent the concentrated shifting of technological upgrade costs onto workers and society.

(4) Promote the reasonable feedback of technological benefits to employment, and improve mechanisms for enterprises and society to jointly share transformation costs. The priority directions for technology application should include replacing dangerous positions, reducing labor intensity, improving personnel safety, and assisting professional judgment, rather than merely reducing labor input. It is recommended to encourage enterprises to prioritize using AI to improve working conditions, enhance work safety and service efficiency, retain necessary human-machine collaboration positions, and allocate some of the efficiency improvement gains toward employee skill development, wage improvements, and job creation. Support leading AI enterprises, industry associations, and public welfare organizations in legally establishing advanced technology sustainable development public welfare funds, guiding enterprises to use voluntary donations and matching charitable contributions to allocate part of their technological development gains toward retraining workers affected by AI substitution, supporting youth employment, and enhancing digital skills for disadvantaged groups. Encourage enterprises to use part of the gains formed by improved production efficiency for employee skill development, wage improvements, and job creation, and incorporate employment assistance effectiveness into the evaluation of public welfare projects. This promotes the reasonable extension of technological dividends from within enterprises to workers and society, forming a socially shared and co-responsible mechanism that coordinates technological development benefits with employment transformation costs.

(Chen Bing is the Director of the Competition Law Research Center and Deputy Dean of the Law School at Nankai University; Ni Lebing is an Assistant at the Competition Law Research Center, Nankai University)

Source

Yicai