AI agents are far less efficient than humans, so why are tech companies still laying off staff? The truth behind the efficiency paradox

A bizarre "efficiency paradox"

On one hand, test reports show AI agents are riddled with errors and far less efficient than humans; on the other, tech companies are still conducting massive layoffs. A recent in-depth report by CBC reveals the cracks in this contradiction: when AI can't even handle simple customer service, why are companies still using "AI replacement" as a reason to optimize their workforce? Behind this lies not a sign of technological maturity, but a shift in corporate decision-making logic — from "efficiency first" to "cost and narrative first."

Core facts: AI agents underperform, but layoffs continue

In May 2026, CBC cited multiple studies showing that current mainstream AI agents have a success rate of less than 30% on complex tasks, far below the 80%+ rate of human employees. For example, in scenarios requiring multi-step reasoning such as customer service and information verification, AI agents frequently "hallucinate" wrong answers and miss critical steps. Yet, according to the tech layoff tracker Layoffs.fyi, global tech companies laid off over 160,000 people in 2025, and more than 80,000 in 2026 so far, with a significant portion of layoffs directly or indirectly citing "AI replacement" as the reason.

Other related developments in the same period:

• Qwen App announced full opening to third-party agents, with brands like Luckin Coffee and KFC testing AI agent services (36Kr, 2026-06-03).
• Kling AI split off and completed its first financing round at a valuation of $18 billion, planning a Hong Kong IPO in 2027 (36Kr exclusive).
• Volcano Engine raised its MaaS revenue target to 15 billion yuan, compared to only 1.5 billion yuan actual in 2025 (36Kr exclusive).
• ChatGPT reached 1 billion global users, becoming the fastest app to achieve that milestone (PYMNTS).

These figures show a stark "efficiency paradox": AI is being feverishly invested in, while its actual capabilities fall short.

Plain-language breakdown: Since AI is "dumber," why do companies still fire people?

Think of it this way: You run a restaurant and hire 10 chefs, each capable of cooking 80-point dishes. Now a tech company pitches you an "AI cooking robot," claiming one robot equals five chefs, but in tests it often burns the food, uses wrong ingredients, and ends up only good as a side dish. Yet to prove to investors that you're "embracing AI," you still fire half your chefs and buy two robots. The result: customer complaints about food quality drop, but the salary savings just cover the robot costs — that's today's "efficiency paradox."

Corporate decision-makers are well aware of AI's shortcomings, but three forces push them toward "irrational" choices:

Driving Factor	Specific Manifestation	Underlying Logic
Capital Narrative	Companies that don't lay off are seen as "backward" by Wall Street	Stock prices and fundraising need an AI story
Cost Reduction	Even if AI is inefficient, labor costs are higher (especially for customer support, data labeling)	Better short-term financial reports
Technological Optimism Bet	Belief that AI capabilities will soon improve exponentially, so they position early	Seize the ecosystem niche, better to overkill than be left behind

This is not a technology problem, but an economics and decision psychology problem.

Impact by group: Who gets hurt, who benefits?

Professionals (especially customer service, content moderation, junior programming)

• Benefit: Those skilled in AI tools become more sought after — companies need "AI tamers" to patch up agent flaws.
• Risk: Undifferentiated entry-level roles may shrink, but full replacement is still a ways off.
• Fad or not? Not recommended to quit your job to "learn AI" just yet, but improving problem decomposition and cross-domain communication skills is advised.

Students / New graduates

• Benefit: Emerging roles like "AI trainer" and "prompt engineer" are in growing demand.
• Risk: Traditional paths (e.g., coding, customer service) have narrowing windows of opportunity.
• Fad or not? Prioritize solid fundamentals over chasing trending tools.

Creators / Self-media

• Benefit: AI assistance in drafting and image generation can boost efficiency.
• Risk: A flood of low-quality AI content impacts original recognition and revenue.
• Fad or not? Use AI like an "intern" — but handle key creativity and fact-checking yourself.

Ordinary users

• Benefit: Some services become cheaper (e.g., smart customer support).
• Risk: Encountering AI "hallucinations" leading to incorrect information (e.g., medical advice, legal counsel).
• Response: Stay skeptical of AI-generated content; actively seek human confirmation for critical matters.

Neutral pros & cons + pitfalls: Don't fall for "replacement anxiety"

Advantages

• AI does excel at repetitive, standardized tasks (e.g., simple FAQs, data extraction).
• In the long term, technological advances may fix current shortcomings.

Disadvantages

• Current AI agents lack common sense and cannot handle novel scenarios; human oversight remains essential.
• Ethical and legal risks: An IBA report notes that AI used in hiring can produce discrimination (e.g., implicit bias based on gender, race).
• Health sector risks: A 2026 survey shows doctor-patient concerns over AI diagnostics center on transparency and liability.

Pitfall avoidance guide

• Don't believe the "full AI replacement" hype: Most promoters are selling anxiety, not practical solutions.
• Use AI cautiously for critical decisions: Fields like law, health, and finance require human review.
• Don't blindly pay for expensive AI courses: Many are outdated and impractical; prioritize free resources with hands-on projects.

Light humanistic reflection: The boundaries of efficiency and human irreplaceability

Ancient Greek philosopher Aristotle once said: "Tools are means to an end, but the human being is an end in itself." No matter how powerful AI becomes, it remains a means. When companies preemptively fire humans for the sake of an "efficiency narrative," we lose not just jobs but a kind of social resilience built on imperfect, warm collaboration.

History repeatedly shows that during technological transitions, the first to be eliminated are often not "those who can't use AI," but those who don't understand their own core value. Your empathy, creativity, and cross-domain integration ability — these very "shortcomings" of AI — are precisely your moat.

Light interactive question

Have you encountered a real case of someone being laid off due to "AI replacement"? Or, after deploying AI in your own work, did efficiency improve or backfire? Feel free to share your observations in the comments as we unpack the efficiency paradox of our time.