WAIC 2026: AI Enters the Workforce – Will Workers Be Replaced or See Wage Increases?

Executive brief

Core judgment: Under the base case, workers' jobs will not disappear but shift to a human-AI collaboration model. However, wage growth and job retention are highly dependent on whether corporate training investment and institutional safety nets are implemented synchronously; individual efforts alone cannot achieve a smooth transition.

Stance landscape: All parties have reached a strong consensus on the direction of human-AI collaboration with no opposing voices, yet significant interest tensions exist regarding who bears the transition costs and whether technological maturity matches the window of opportunity.

Maximum risk: If the status quo of technology leading while training lags persists, the gap between workers' existing skills and new job requirements will trigger transitional unemployment. Macro-level collaboration consensus is prone to alienating into structural pain borne by individuals at the micro level.

Turning-point signals: Closely monitor the signing volume and cost curve trends of mainstream service providers' solutions for SMEs, while also paying attention to the proportion of AI replacement disputes in labor arbitration and abnormal fluctuations in resignation rates.

Recommended actions: Immediately promote the establishment of an institutionalized mechanism for sharing transition costs. Require enterprises to simultaneously implement supporting training budgets and define safety responsibilities when procuring AI equipment, avoiding placing the trial-and-error costs solely on workers.

This Executive brief summarizes the conclusions of the full report; see corresponding chapters in the main text for the sources and basis of each judgment.

Distribution of Stance among Key stakeholders: Conditional support under consensus

This analysis centers on the proposition that "workers should proactively learn new skills to transition into a human-AI collaboration model." The World AI Conference (WAIC) 2026 was held in Shanghai from July 17 to 20. Its theme, "Intelligent Partner," marks the AI industry's shift from proof-of-concept to practical implementation (Source: material). The simulation covers 15 key stakeholders, including national ministries, manufacturing enterprises, technology vendors, investment institutions, frontline practitioners, and ordinary families, generating a total of 98 valid stance posts (excluding observers). Statistics show that 87.8% of the posts support the proposition, while 12.2% hold a neutral, wait-and-see attitude, with no opposing voices (Source: simulation). Overall public opinion exhibits the characteristic of "strong consensus on direction but stringent conditions for implementation": all parties generally agree that human-machine collaboration is an inevitable trend, but they unanimously emphasize that the costs of transformation cannot be borne by workers alone; it must be accompanied by training systems, safety baselines, and job reconstruction mechanisms, otherwise "proactive learning" will remain empty talk.

Technology and Platform Providers: Exchanging low barriers and safety baselines for collaboration space. Simulation roles based on Alipay and WeChat AI clearly define the safety boundary of "AI only handles errands, while fund decision-making power remains with users," believing that lowering usage barriers is the prerequisite for people to be willing to team up with AI (Source: simulation). A simulation role based on StepFun further points out that agent products must be "ready-to-use upon opening," and enterprises providing equipment must accompany it with training; this is a basic responsibility, not an extra favor (Source: simulation). A simulation role based on Moonshot AI focuses on the supply side, emphasizing that the value of open-source models lies not in the code itself, but in whether supporting deployment tools, documentation systems, and industry adaptation solutions enable ordinary business personnel to truly use them (Source: simulation). The common logic of these three types of roles is: technological inclusivity does not equal technological laissez-faire; only when tools are sufficiently easy to use, safety boundaries are clear, and learning costs are controllable does "proactive learning" become realistically feasible.

Manufacturing Enterprises and Frontline Practitioners: Supporting collaboration but requiring synchronized training. A simulation role based on BYD admits that robots entering production lines aim to supplement labor shortages rather than replace humans, but the pace of transformation must follow production rhythms; production lines cannot be stopped for training, and enterprises should bear the majority of training costs (Source: simulation). A simulation role based on Micro-E Zone adds that equipment launch is merely the starting point; phased training allowing experienced workers to gradually master operational logic is key to implementation, emphasizing that "transformation costs cannot be dumped on individuals" (Source: simulation). A frontline pharmacist role draws a professional bottom line: repetitive sorting can be handed over to machines, but prescription review authority and safety responsibilities must be controlled by humans (Source: simulation). These voices indicate that industry support for human-machine collaboration is conditional—the condition being that job reconstruction and skill training must advance synchronously with technological deployment, and core decision-making power cannot be ceded.

Ordinary Workers and Families: Acknowledging the direction but anxious about cost-sharing. The worker persona expresses a complex mindset of "panic yet acceptance" multiple times during the simulation: acknowledging the fact of AI efficiency gains, but more worrying that companies will "only deploy equipment without providing training," pushing tuition and time costs entirely onto individuals (Source: simulation). An ordinary viewer family persona Speak plainly ly states, "With elderly parents above and young children below, where is there time and energy to spend money out-of-pocket to learn new skills?" calling for ordinary families not to bear the entire transformation cost (Source: simulation). The stance of such roles is not opposition to learning, but a strong demand regarding "who pays, who organizes, and who guarantees." Their anxiety points to a core contradiction: if macro-level "should proactively learn" lacks micro-level institutional support, it easily degenerate into moral kidnapping of vulnerable groups.

Policy and Academic Sectors: Emphasizing systematic backing and employment security. A national ministry role explicitly states during the simulation that it will "improve the vocational skill training and employment security system, firmly holding the bottom line of preventing large-scale unemployment" (Source: simulation). A simulation role based on Liu Tieyan points out from an educational perspective that the key to collaboration rather than replacement lies in whether systematic education can keep up, and that technological vitality depends on the connection between open mobility and talent cultivation (Source: simulation). A WAIC organizer role summarizes that the consensus among all parties points to "AI implementation is not replacement but collaboration, and requires simultaneous advancement of capital, technology, training, and institutions" (Source: simulation). These statements provide the institutional legitimacy premise for "proactive learning": individual efforts must be embedded within a public support network, rather than facing technological shocks in isolation.

Key points of divergence and key uncertainties: cost allocation and data bottlenecks

The simulation suggests that while there is broad consensus on the direction of workers proactively learning new skills to transition into a human-AI collaboration model, two core divergences remain regarding implementation: who bears the cost of transition and whether technological maturity aligns with the transition window. These divergences do not oppose collaboration itself but are key variables determining whether the collaboration model will be a smooth transition or trigger structural friction.

Core Divergence: Institutionalized Allocation of Transition Costs vs. Individual Bottom-Line Responsibility

In the Simulation, all stakeholders agree on the necessity of human-machine collaboration, but significant interest tensions exist regarding "who pays for learning new skills." Frontline practitioners and ordinary families generally expressed anxiety about "fear of being eliminated but inability to afford self-funded transition," emphasizing that enterprises must provide simultaneous training when introducing AI equipment; otherwise, "proactive learning" will mutate into moral kidnapping of laborers (Source: simulation). In contrast, Simulation characters based on Weiyi Intelligent Manufacturing and Ant Lingbo, although promising supporting services such as "half-day onboarding" and "phased mentoring," base their statements more on the logic of product landing and commercial closure—that "if no one can use it, the solution has not landed"—rather than pure public welfare responsibility (Source: simulation). Simulation characters based on Qiming Venture Partners propose from a capital perspective that training budgets should be included as "rigid expenditures" in enterprise AI infrastructure procurement, implying that without institutional constraints, market spontaneous mechanisms struggle to ensure fair distribution of training resources (Source: simulation). Verified online materials show that fees for enterprise-specific AI system courses range from tens of thousands to hundreds of thousands of RMB per tier, and require subsequent consulting services to ensure effectiveness (Source: verified online, url: nabi.104.com.tw), further confirming that without institutional support from enterprises or public funds, ordinary workers alone lack the financial capacity to bridge the skills gap. The current divergence essentially lies in: Is training an attached gift for equipment sales, or an independent right regulated by policy? If the latter is not established, the inclusiveness of the collaboration model will face severe challenges.

Key Uncertainty 1: Speed of Accumulation of High-Quality Scenario Data

Technological maturity directly determines the "safety window period" for worker transition. Simulation characters based on Qiming Venture Partners point out that high-quality physical interaction data from leading embodied AI companies in China and the US is currently at the level of several hundred thousand hours, still short of one order of magnitude from verifying Scaling Law, which is the core bottleneck restricting AI from "being able to run" to "being able to work" (Source: simulation). This means full AI replacement of complex manual labor will take time, but also brings uncertainty: If data bottlenecks are broken through slower than expected, enterprises may reduce AI investment due to ROI falling short of expectations, leading to insufficient supply of "human + AI" jobs; If breakthroughs happen faster than expected, technical iteration speed might exceed the response capability of training systems, causing workers who have just completed transition to face skill devaluation risks again. In the Simulation, Simulation characters based on Ant Lingbo emphasize that "real scenario data run under pressure" has more verification value than laboratory data (Source: simulation), suggesting decision-makers need to focus on data accumulation progress in specific industries rather than generalized technical parameters.

• Signals to watch: Focus on monitoring the duration of continuous stable operation and trends in failure rates of AI systems under real production rhythms in benchmark scenarios such as BYD welding workshops and offline pharmacies, rather than only paying attention to exhibition demonstration effects. Note that the metrics mentioned in materials for Weiyi Intelligent Manufacturing robots, such as "error <0.02mm, zero missed detection rate," have not been traced to original sources (The data has not been traced to original sources); subsequent third-party audit reports or actual production line test data published by enterprises should prevail.

Key Uncertainty 2: Balancing Reduction in Reasoning Costs with Deployment Flexibility

Even if data bottlenecks are broken through, economic feasibility remains another variable determining whether the "human + AI" model can be popularized among SMEs. Simulation characters based on StepFun mention that reducing reasoning costs and improving deployment flexibility are prerequisites for agents embedding into office scenarios (Source: simulation); Simulation characters based on Alipay/WeChat AI also admit that if reasoning costs cannot be reduced, scaling up is empty talk (Source: simulation). For the vast number of small and medium-sized manufacturing and service enterprises, only when AI usage costs fall below the comprehensive cost of "labor + training" will business owners have the motivation to retain and upgrade existing positions rather than simply laying off and replacing staff. Currently, the debut of Huawei Atlas 950 SuperPoD and Sugon 8000 super-cluster shows progress in autonomous and controllable computing power foundations (Source: material), but there is a transmission lag between computing power supply and the reduction of terminal application costs; the length of this lag will directly affect the width of the transition window period.

• Signals to watch: Track API call prices, equipment rental rates, and actual signing volumes of "out-of-the-box" solutions launched by mainstream AI service providers targeting SMEs, observing whether the cost curve enters a rapid downward channel.

Risk warning: Structural friction during the transition pain period

If the human-machine collaboration transformation proceeds under the current status quo of "technology first, training lagging," it may trigger three types of structural friction risks in the short term. These risks do not stem from AI technology itself, but from the time lag between institutional support and technology implementation. Without early intervention, a macro-level "collaboration consensus" is highly likely to be alienated at the micro-level into the transitional pains borne by individuals.

Risk of "transitional unemployment" triggered by skill mismatch

Although WAIC 2026 demonstrated AI's Implement capabilities in scenarios such as production line quality inspection and pharmacy sorting, the gap between workers' existing skills and the requirements for AI collaboration roles has not been effectively bridged. In the simulation, characters modeled after office workers repeatedly expressed anxiety about "companies buying equipment without teaching people how to use it" and "veteran employees being sidelined" (Source: simulation). This anxiety is supported by real-world data: according to a report in Sichuan Daily, the iteration cycle of AI technology has been compressed to a "monthly" level, with nearly half of digital roles requiring skill updates every six months. The insufficient digital competence of traditional practitioners exacerbates the structural contradiction of coexisting "difficulty in finding jobs" and "difficulty in recruiting" (Source: verified online, url: epaper.scdaily.cn). If enterprises view AI solely as a cost-reduction tool rather than an opportunity for workforce upgrading, and lack supporting job reconstruction and paid training mechanisms, a large number of middle-aged and low-skilled workers may be forced to exit the labor market under the moral pressure of "active learning," forming a group of technically unemployed individuals.

Trap of "pseudo-collaboration" caused by cost shifting

The lack of a mechanism for sharing transformation costs is another major hidden danger. The simulation shows that although technology vendors promise "plug-and-play" and supporting training, characters modeled after corporate administrators point out that in actual procurement, "training budgets" are often stripped or compressed, leading to idle equipment or employees paying for their own learning (Source: simulation). When "human+AI collaboration" becomes merely a slogan, and enterprises do not include training as a rigid expenditure, the so-called "collaboration" may turn into disguised performance exploitation—employees must bear the implicit labor of adapting to new systems free of charge in addition to completing their core duties. An article from the World Economic Forum warns that sudden unemployment caused by AI brings not only income loss, but also identity crisis and loss of social belonging; this psychological burden is severely underestimated in current policy discussions (Source: verified online, url: cn.weforum.org). Without institutional guarantees for "collaboration quality," technological dividends may be monopolized by those who master AI tools, exacerbating income differentiation and social instability.

Risk of responsibility vacuum due to the breach of safety bottom lines

In the absence of legally defined boundaries for human-machine collaboration, ambiguous zones of decision-making authority and responsibility attribution may lead to accidents. In the simulation, Alipay WeChat AI and Ant Lingbo both emphasized the safety bottom line of "AI only handles Run errands, keeping decision-making power with humans" (Source: simulation), and national ministries have reiterated the principle of "humans as the main body, AI as auxiliary" (Source: simulation). However, in practice, when AI systems provide incorrect advice or execute deviations, if enterprises fail to establish clear manual review processes and responsibility tracing mechanisms, frontline employees may be forced to take the blame for algorithmic errors. Especially in high-risk fields such as healthcare and manufacturing, if key responsibilities such as prescription review authority and quality signing authority are tacitly ceded to AI, it may breach regulatory red lines. In response to the risk of AI-induced unemployment, the Ministry of Industry and Information Technology also emphasized the need to improve the employment security system to cope with technological shocks (Source: verified online, url: finance.people.com.cn), suggesting that the current institutional framework still has gaps in covering new types of human-machine responsibility relationships.

These risks indicate that the validity of the proposition that "workers should actively learn" highly depends on the simultaneous availability of external support systems. If relying solely on individual efforts without enterprise training investment, cost-sharing mechanisms, and clear safety responsibilities, the transition period may slide from "smooth transition" to "structural rupture," ultimately undermining the sustainable development of the AI industry itself.

Recommended actions: Skills transition pathways for scenario-based responses

The decision recommendation should go beyond calling on individuals to "proactively learn" and instead build a skill transition support system that is scenario-based, executable, and has safety nets. This is based on the "smart partner" signal released at WAIC 2026 and the strong consensus among all parties in the simulation regarding the allocation of transition costs. The following recommendations provide differentiated action guidelines for different development paths around the proposition that workers should proactively learn new skills to transition into a human-AI collaboration model.

1. Base Case: Institutional Support and Technology Implementation Advance Simultaneously

• Trigger conditions: Steady growth in signed volumes of "plug-and-play" solutions for SMEs by mainstream AI service providers, and the proportion of training budgets paired with enterprise AI equipment procurement reaching industry guidance standards; the conversion rate of Intent cooperation projects reached at WAIC 2026 (according to authoritative media reports such as Sina Finance, the expected Intent procurement amount is approximately RMB 20.36 billion, which deviates from the material's original expression of RMB 16.2 billion) exceeds 30% within half a year. (Source: verified online, url)
• Key actions: The Ministry of Human Resources and Social Security jointly with the Ministry of Industry and Information Technology issues the "Guidelines for Vocational Skills Training in the AI Era," clarifying that enterprises must include training as a mandatory expenditure when introducing AI equipment, and establish a filing mechanism of "training must be equipped upon equipment launch, and plans must exist for position adjustments." Vocational colleges and community education centers collaborate with technology vendors to develop modular courses, focusing on two skill stacks: blue-collar AI operations and maintenance, and white-collar AI scheduling, referencing open-source knowledge bases like WaytoAGI to lower learning thresholds. (Source: verified online, url) Ordinary workers should prioritize mastering the "supervisor" capability of AI tools in their own positions, i.e., understanding AI output logic, identifying abnormal signals, and retaining manual review rights, rather than pursuing full-stack technical capabilities.
• Signals to watch: Track the data on applications for AI skills training subsidies issued by local human resources departments, the coverage rate of training clauses in enterprise equipment procurement contracts, and quarterly survey scores from frontline practitioners on "human-machine collaboration comfort."

2. Best Case: Low-Cost Inclusive Tools Accelerate Skill Equity

• Trigger conditions: Domestic computing power bases (such as Huawei Atlas 950 and Sugon 8000) drive inference costs down by more than 50% within one year; mature localization deployment toolchains for open-source models (such as Kimi K3) enable non-technical personnel to complete business adaptation in a no-code environment; high-quality physical interaction data for embodied intelligence breaks through to the million-hour level, validating the effectiveness of Scaling Laws in real-world scenarios. (Source: material)
• Key actions: Technology platform providers should treat "document localization, industry adaptation solutions, and age-segmented mentoring tutorials" as standard product deliverables rather than Additional services. Enterprise administration and HR departments can use low-cost AI infrastructure to quickly build internal skills training platforms, embedding "learning by doing" into production processes to avoid disconnect between off-the-job training and business operations. Workers can leverage low-threshold entry points of national-level apps (such as Alipay's "A Bao" and WeChat AI) to naturally accumulate AI collaboration experience in daily work, combining soft skills (communication, judgment, empathy) with AI tools to form differentiated competitiveness. (Source: material)
• Signals to watch: Monitor the completeness of Chinese documents for mainstream open-source models, the month-on-month growth rate of SME AI API calls, and the frequency and task completion rate of non-technical employees using AI tools independently.

3. Worst Case: Lagging Institutions Exacerbate Transition Pains

• Trigger conditions: Enterprises generally cut training budgets to compress costs, purchasing only equipment without supporting capacity building; frequent AI application safety accidents lead to tightened regulation, blurring human-machine responsibility boundaries and triggering labor disputes; data bottlenecks for embodied intelligence remain unresolved, with technology implementation stagnating at the Demo stage, unable to create enough new positions to absorb transferred personnel. (Source: simulation)
• Key actions: Regulatory authorities should immediately initiate special assessments of AI's impact on employment, implementing credit constraints or tax adjustments for enterprises that introduce AI on a large scale but fail to fulfill training obligations. Trade unions and industry associations take the lead in establishing a "transition mutual aid fund" to provide transitional living subsidies and retraining opportunities for workers technologically unemployed. Workers need to beware of the "pseudo-collaboration" trap; if enterprises only provide equipment without position reconstruction and safety nets, they should strive for rights protection through collective consultation to avoid bearing transition risks alone. In the simulation, characters prototyped from ordinary audience families repeatedly emphasized the realistic dilemma of "expensive tuition and no connections," highlighting the urgency of policy intervention. (Source: simulation)
• Signals to watch: Pay attention to the proportion of labor arbitration cases involving AI substitution disputes, abnormal fluctuations in resignation rates in manufacturing and service industries, and changes in the sentiment index of topics related to "AI anxiety" on social media.

Regardless of the scenario, the essence of "human + AI collaboration" is that human subjectivity is not dissolved by technology. Decision-makers must keep in mind: the theme of WAIC 2026 is "smart partners" rather than "smart substitution." Wage increases or decreases depend on whether workers can team up with AI, rather than being singled out by it. (Source: material) Only by elevating skill transition from individual struggle to a systematic project can technological dividends truly benefit every worker.

Source materials

Verified online sources

• Complete Guide to Enterprise AI Adoption: AI Course Fees, AI Consultant Fees, AI Customer Service... - AI Learning | 104 Learning
• How to View the Unemployment Risks Posed by AI Development? MIIT Responds -- Economy & Technology -- People's Daily Online
• Building a Resilient Employment System to Address the Risk of "AI Unemployment" --- Sichuan Daily Electronic Edition
• The "Precariat" Caused by the AI Wave is an Overlooked Global Risk | World Economic Forum
• A Comprehensive and Systematic AI Learning Path, Helping Ordinary People Master AI - Chasing Time - Cnblogs

Material Sources

• tfcaijing.com
• <finance.sina.cn
• readmo.cmoney.tw