AI Is Here—Don't Rush to Lay Off Staff! Insights for Businesses from Google’s Chief Economist: Redesigning Work Matters More Than Eliminating Employees

"AI will transform far more jobs than it destroys."

This is the key observation on artificial intelligence and the labor market shared by Google’s Chief Economist, Fabien Curto Millet.

Citing research by MIT economics professor David Autor, Millet points out:

"Roughly 60% of jobs in the US economy today did not exist 80 years ago."

Millet states:

"Technology is generally a net creator of jobs."

For instance, when personal computers proliferated across the United States, industries and positions such as typewriter manufacturing and traditional clerical work lost approximately 3.5 million jobs. Yet that same technological revolution generated roughly 19 million new jobs in areas like computer manufacturing and data analysis.

Citing research by his Google colleague James Manyika, Millet explains that while technological progress renders certain roles obsolete, it also has the potential to spawn far more new employment opportunities. He anticipates AI will follow a similar trajectory.

For example, an emerging role that garnered significant attention last year—the "Forward-Deployed Engineer"—is one of the new job categories catalyzed by AI. These engineers typically work directly with clients or internal enterprise stakeholders to deploy AI technologies into practical operational scenarios, resolving concrete problems.

Millet suggests this may merely be an early signal of AI creating a broader array of novel job profiles in the future. He emphasizes that the most profound productivity gains from AI do not come from merely shoehorning AI tools into legacy workflows, but from fundamentally redesigning how work is done. In the future, managers will need to govern not only human employees, but also AI Agents.

This observation offers profound implications for HR executives and managerial leaders across Taiwan.

When AI can handle procurement, travel coordination, financial modeling, contract review, and customer support, should organizations downsize their workforce or redesign the work itself?

My counsel is: Redesign work first, empower employees to transition and adapt to the AI era second, and evaluate layoffs as an absolute, unavoidable last resort only when all other options fail.

Under Article 11 of Taiwan's Labor Standards Act, deploying AI does not in itself constitute a statutory ground for severance dismissal. An enterprise cannot deem a dismissal lawful simply because AI operates faster and cheaper.

Enterprises should implement the following three concrete steps:

For labor compliance training, delineating reasonable management vs. workplace bullying, grievance handling & investigation simulation workshops, board member continuing education certification, external committee members for workplace bullying and sexual harassment, PIP (Performance Improvement Plan) case simulation workshops, and litigation/mediation representation, please contact: Yesin Law Firm (Tel: +886-2-2515-6822 | Official Website: https://www.yesinlaw.com).

1. First, Redesign Operational Workflows

Suppose a tech company originally employs 20 procurement specialists. Following the rollout of AI Agents, supplier quote comparisons, purchase order generation, delivery tracking, and reporting are fully automated. The company concludes it now needs only 10 procurement specialists and prepares to lay off the other 10.

Is this lawful?

First, we must review the five statutory grounds for unilateral termination under Article 11 of the Labor Standards Act:

1. Subparagraph 1: Cessation of operations or transfer of business.

2. Subparagraph 2: Operating losses or business contraction.

3. Subparagraph 3: Suspension of work for more than one month due to force majeure.

4. Subparagraph 4: A change in the nature of business creating the necessity to reduce the workforce, and no suitable alternative work is available for reassignment.

5. Subparagraph 5: The employee is demonstrably incompetent to perform the assigned duties.

Introducing AI does not inherently satisfy any of these statutory categories.

For example, automating procurement workflows may indeed involve a "change in the nature of business" under Subparagraph 4. However, the enterprise must still substantiate that a substantive operational change genuinely occurred, that reducing headcount is strictly necessary, and that no suitable alternative work exists for reassignment. Merely increasing operational efficiency and trimming overhead does not legally justify severance dismissals.

I advise HR to establish an "AI Workflow Redesign Evaluation Matrix" to audit:

• Which tasks can be automated by AI?

• Which tasks strictly require human judgment?

• Which roles can be restructured and combined?

• Which employees can be redeployed to alternative functions?

For example, staff previously tasked with inputting purchase orders can transition into supplier risk management, AI procurement oversight, or strategic sourcing roles.

True organizational transformation is not about slashing 20 employees down to 10; it is about figuring out how those 20 employees can deliver the value previously generated by 40.

2. Next, Upskill Employees in AI Capabilities

An employee’s unfamiliarity with AI tools cannot simply be equated with "incompetence to perform the job."

Google’s Chief Economist specifically emphasizes that employee retraining and upskilling represent non-negotiable investments in AI transformation. This aligns seamlessly with the "Principle of Dismissal as a Measure of Last Resort" (解僱最後手段性原則) consistently upheld by Taiwan's labor courts.

While Article 11, Subparagraph 5 of the Labor Standards Act allows termination with advance notice when an employee is demonstrably incompetent to perform their duties, an employee’s lack of familiarity with new AI tools does not automatically constitute statutory incompetence. In particular, if an enterprise abruptly deploys a new system without providing adequate training, a reasonable learning runway, and necessary tools, and then fires employees for poor performance, it faces immense legal exposure.

I advise enterprises to implement an "AI Skill Transformation Roadmap":

• Step 1: Skill Assessment & Diagnosis: Map existing capabilities, identify how AI relates to each role, and provide foundational training.

• Step 2: Hands-On Application: Facilitate practical application of AI tools under managerial mentorship, establishing reasonable and measurable performance benchmarks.

• Step 3: Performance Appraisal & Role Realignment: Assess whether employees have developed the competencies required for emerging roles. If gaps persist, initiate a formal Performance Improvement Plan (PIP) and evaluate alternative internal transfer possibilities.

Furthermore, employers must heed the anti-discrimination mandates under Article 5 of the Employment Service Act and the prohibition against age discrimination under Article 12 of the Middle-aged and Senior Employment Promotion Act.

Never assume older employees are ill-suited for the AI era simply because they are older or take longer to adapt. Training only younger workers while earmarking employees over 45 or 50 as priority targets for layoffs invites severe age discrimination liabilities. AI proficiency must be evaluated against objective criteria, not based on age or managerial bias.

3. Finally, Evaluate Workforce Adjustments

Severance dismissals must rest upon solid statutory grounds backed by robust documentation.

Even after redesigning workflows, delivering training, and exhausting internal reassignment options, situations may still arise where specific positions genuinely vanish or individual workers prove unable to adapt to new requirements. When this occurs, HR must carefully differentiate between distinct legal scenarios:

• If organizational transformation drives a bona fide change in the nature of business, headcount reduction is unavoidable, and no suitable alternative roles exist for redeployment, Article 11, Subparagraph 4 of the Labor Standards Act may apply.

• If the company experiences actual operating losses or business contraction, assessment must be made under Article 11, Subparagraph 2 based on concrete financial facts—not speculative future payroll savings from AI.

• If an employee remains incompetent after receiving reasonable training, coaching, and a structured opportunity to improve, Article 11, Subparagraph 5 may apply, provided it satisfies objective reasonableness and the principle of dismissal as a measure of last resort.

I advise organizations to maintain an "AI Transition Workforce Reassignment Record" preserving at least five categories of documentation:

1. Pre- and post-AI job descriptions and structural adjustment records.

2. Employee competency assessments and retraining documentation.

3. Performance evaluations, mentoring logs, and PIP progress reviews.

4. Internal job vacancy postings, transfer consultation notes, and redeployment evaluations.

5. Final workforce adjustment determinations and supporting evidence for statutory grounds of termination.

If severance becomes truly unavoidable, the employer must strictly fulfill advance notice periods, file severance notifications, disburse statutory severance pay, and issue Certificates of Involuntary Separation, while verifying whether the thresholds under the Act for Worker Protection of Mass Redundancy are triggered.

AI transformation must never serve as a pretext to unilaterally reduce wages, coerce resignations, or evade severance liabilities.

Attorney Chen Yeh-hsin's Key Takeaway

In the age of AI, the most vital managerial capability is not cutting headcount—it is enabling employees to transform successfully.

The observations from Google’s Chief Economist remind us of a critical truth: AI’s greatest value is not merely replacing humans in existing tasks, but unlocking novel ways of working that were previously unimaginable.

For businesses, AI transformation must follow a strict three-stage sequence:

Redesign work first, invest in employee capabilities second, and evaluate workforce adjustments in accordance with the law last.

This is not merely corporate social responsibility; it is a pragmatic operational strategy to mitigate labor disputes, safeguard employer brand equity, and build enduring competitive advantage.

I frequently remind corporate leaders: a comprehensive training record, a sincere transfer consultation, and an authentic PIP offering a genuine opportunity to improve can serve as vital legal evidence proving the company exhausted its duty of reassignment and honored the principle of dismissal as a measure of last resort.

More importantly, however, these measures may allow the company to retain exceptional talent—employees who embody institutional knowledge, bring years of practical domain experience, and are eager to evolve alongside AI.

In the age of AI, the most valuable investment is not just smarter artificial intelligence tools, but the people who know how to wield them.

The rise of AI is unstoppable. HR leaders should immediately audit ongoing AI deployments across the company to verify whether affected employees have been granted opportunities for retraining, role redesign, and reasonable reassignment.

When your organization prepares to replace a role with AI, have you genuinely explored how to first use AI to help that employee generate even greater value?