Artificial Intelligence and Employment: The Chinese Xinyuan Case and the Lessons for Europe

A recent ruling from the Shanghai Labour Court has reignited global debate over whether the adoption of artificial intelligence (AI) can legally justify employee dismissals. The court held that the replacement of human workers by an AI system does not, by itself, constitute legitimate grounds for termination, ordering Xinyuan Technology Co., Ltd. to reinstate a dismissed employee and pay compensation for unlawful dismissal.

This decision, although rooted in the Chinese legal system, resonates far beyond its borders. It invites a reconsideration of how technology-driven rationalizations align with employment protections in other jurisdictions — particularly in Europe, where labour law still stands as a counterbalance to efficiency-led restructuring.

The Xinyuan Technology Case: Facts and Legal Reasoning

The case concerned an administrative employee dismissed in 2024 after Xinyuan Technology implemented an internal AI solution that automated routine tasks such as supplier communication and data reporting. The company argued that digital transformation had rendered the position redundant.

The employee challenged the dismissal, invoking Article 40(3) of the Chinese Labour Contract Law, which allows layoffs due to “substantial changes in economic or technological circumstances” only when reassignment or retraining is impossible.

The Shanghai Court found that Xinyuan had neither proven such impossibility nor taken steps to redeploy or retrain the worker. The use of AI, the judgment stated, “represents a tool for business process enhancement, not a legitimate replacement of the human role unless organizational necessity is duly verified.” The ruling thus framed AI adoption as a managerial choice subject to proportionality and social responsibility, not an unconditional right of the employer.

Can an AI Agent Truly Replace Human Labour?

The appeal of AI often lies in the promise of efficiency and cost reduction. But the real economics tell a more complex story — especially when comparing different labour markets.

Average European Scenario

In the European Union, the total cost of employment (gross salary plus employer contributions) for an administrative or knowledge-worker profile typically ranges between €40,000 and €70,000 per year. For more technical roles — such as customer-support analysts, paralegals, or junior data specialists — costs can reach €80,000–100,000 in high-cost countries like France or the Netherlands.

A customised AI agent, trained on corporate data and integrated into internal workflows, often requires:

• €50,000–€200,000 in initial development or licensing

• €1,000–€5,000 per month for hosting, maintenance, and compliance

• Additional human oversight from IT, data-protection, or legal teams

While the variable cost per task may drop, total operational costs often remain comparable once supervision, liability management, and retraining needs are included.

The Swiss Case

In Switzerland, labour costs are substantially higher but also reflect greater productivity and autonomy.

• A qualified administrative or legal staff member generally costs an employer CHF 80,000–130,000 per year, including social charges.

• For specialist or semi-technical roles (compliance officers, IT coordinators, or financial analysts), the total annual cost frequently exceeds CHF 150,000–180,000.

Given these figures, it’s true that automation seems financially attractive at first glance. But most Swiss employers soon discover that an AI system cannot replace the qualitative contribution of an experienced employee — especially in highly regulated or client-facing roles, where judgment, contextual awareness, and ethical accountability are indispensable.

Moreover, AI deployment in Switzerland is subject to data protection (FADP), transparency, and human oversight obligations, which introduce additional compliance costs absent from a purely financial comparison.

Comparative Legal Perspectives: Switzerland, France, and Italy

Switzerland

Swiss labour law, while flexible, prohibits abusive dismissals under Article 336 of the Code of Obligations. Replacing a worker purely for automation reasons may be considered abusive if not rooted in genuine organisational necessity or if the employer neglects the duty of requalification. The principle of good faith (Article 2 of the Civil Code) requires employers to conduct structural changes in a socially responsible way. Emerging discussions also call for alignment with the EU AI Act on transparency and human oversight.

France

Under Article L1233‑3 of the Labour Code, economic dismissals are valid only when driven by a verified "technological change" that genuinely removes the position. Courts insist on proof that redeployment was impossible. If a company uses AI merely to reduce personnel expenses, the dismissal can be deemed void. France also imposes algorithmic transparency obligations for workplace decision-making, reinforcing the worker’s right to contest opaque automation.

Italy

In Italy, dismissals for “objective justified reason” (giustificato motivo oggettivo) are lawful only if a position is truly eliminated for demonstrable economic or technological reasons. Courts interpret Law 604/1966 in light of Article 41 of the Constitution, ensuring that economic freedom remains consistent with human dignity. Italian judges traditionally scrutinize employers’ claims of technological necessity, favouring solutions involving human reallocation or hybrid cooperation with AI tools.

Rethinking Efficiency: A Legal View on Human Value

What emerges across these systems — from Zurich to Paris to Milan — is the same conclusion echoed by the Xinyuan decision: technological innovation cannot justify dismissals unless proportional, transparent, and socially responsible.

AI can streamline repetitive tasks, but it does not yet replicate the ethical reasoning, contextual judgment, and adaptability that define human contribution. In legal and compliance-sensitive roles, these qualities are not optional — they are the ground of trust.

The real challenge for employers, therefore, is not whether AI can replace human labour, but how to integrate both within a compliant and sustainable governance model.

Can an AI agent truly replace a human? Economically, perhaps partially.
Legally and ethically, the answer remains — and should remain — uncertain.