In-Depth Analysis: Why Are Workers in Big Companies More Exhausted When AI Becomes the “New KPI”?
Hello everyone, I’m your financial journalist. Recently, a report about “AI and the workforce in big companies” went viral, and the phenomenon it describes is quite thought-provoking: Everyone says AI can improve efficiency, so why do people feel more exhausted? Some even go out of their own pocket to subscribe to AI services or create “AI versions” of their bosses to meet performance metrics.
This is not just a complaint about the workplace; it reflects the anxiety, alienation, and redefinition of values within the entire internet industry amidst the AI revolution. Today, we’ll break down the core logic of this report into five key points in simple language to understand what’s really happening in this “AI arms race.”
---
1. From “Voluntary Use” to “Mandatory Assessment”: AI as the New “Overtime Pay”
Previously, we competed by working extra hours and meeting KPIs; now, we’re competing by using a certain amount of “Tokens” (units of AI usage).
There’s a typical example in the report: Xiaoqi, a product manager at ByteDance, wakes up several AI assistants every day as her first task. What used to be a tool for laziness has become a necessity for survival. Why? Because bosses are now keeping an eye on it.
- Previously: Whether you used AI was up to you, and using it was a bonus.
- Now: The company provides you with a quota (e.g., 7,000 yuan/month at Tencent, 8,000 yuan/month at Alibaba), and they even cover the cost of external subscriptions. If you use less, the boss asks, “Why are you using so little? Are you not making use of AI?” If you use more, the boss asks, “Where’s the output for all this money?”
This creates a “dual bind”:
1. You can’t avoid using it: Not using AI makes you seem “outdated” or “unwilling to embrace new technology,” which can affect your performance and promotions.
2. Using it recklessly is also not okay: The company starts to track your usage. Tokens aren’t free; they represent real costs. If you spend a million tokens but don’t generate corresponding business value, you’re wasting company resources.
In simple terms: It’s like the company gives you a luxury car (an AI tool) and also provides you with a fuel card. Before, no one cared if you drove it or not; now, if you don’t drive, your boss thinks you’re lazy. If you drive every day, your boss checks your fuel consumption and asks why you haven’t transported more goods. AI has gone from a benefit to a performance metric, and workers not only have to work but also prove they’re working “correctly.”
---
2. Major Shift in Hiring Standards: Without AI Skills, You Don’t Even Qualify for Interviews
The report mentions that over 80% of job openings at companies like Meituan, Ant Group, and Ctrip require AI proficiency. This sends a strong signal: “Knowing how to use AI” is no longer just a bonus; it’s become a prerequisite, even a prerequisite for entry.
Previously, we talked about “versatile talents”; now, it’s “professional skills + AI knowledge.”
- Operations roles: You used to need to write copy and create visuals; now, you need to write prompts and use agents to collect information and generate drafts automatically.
- R&D roles: You used to need to code; now, you need to “direct” AI to write code. If you’re still coding manually, you’re not only slow but may also be replaced by new hires or AI.
In simple terms: The job market is like getting a driver’s license. Before, you needed to know how to drive a manual car; now, you need to know how to use autonomous driving systems. If you only know how to drive manually, you’re considered an inefficient asset in companies that prioritize efficiency. Your competitiveness lies not just in how much professional knowledge you have but in your ability to communicate that knowledge to AI and have it execute tasks for you.
---
3. The Efficiency Paradox: Where’s the Time Saved? Why Are We More Exhausted?
This is the most confusing point: AI is supposed to save us 8 hours of work (according to BCG research), so why do we feel more exhausted?
The answer lies in the “Parkinson’s Law” applied to AI: Work tends to expand until it fills all available time.
1. Standards have risen: Writing a report used to take 3 days; now, AI can do it in 1 hour. The boss doesn’t give you those 2 days off but says, “Since it can be done in 1 hour, you need to submit 10 reports this month.” Or, “If you code so fast, why hasn’t the project been launched yet?”
2. New bottlenecks have emerged: AI solves the “execution” problem, but not the “decision-making” and “collaboration” ones.
- The code is written, but is the requirement correct?
- The report is generated, but are the data figures accurate?
- The plan is ready, but has the boss approved it?
- The new bottleneck is “thinking” and “interpersonal relationships.” You spend more time verifying AI’s outputs, coordinating resources across departments, and explaining why one plan was chosen over another.
In simple terms: Before, you were a bricklayer; now, you’re a supervisor. Being a supervisor is more exhausting because you have to oversee every detail and deal with stakeholders. AI reduces the time for physical tasks but increases the time for mental and communication efforts.
---
4. Absurd “Experiential Techniques”: Using AI to Deal with AI and Bosses
The most interesting part of the report describes workers coming up with bizarre strategies to meet AI-based performance assessments, reflecting the alienation in workplace relationships.
1. **Creating an “AI Boss”: Algorithm engineer Kobayashi fed all the boss’s chat records and decision-making styles into an AI to create a “boss clone.” Before reporting, he practiced with the AI boss to predict what the boss would say and adjusted his speech accordingly, resulting in the highest performance rating.
- Essence: This isn’t about improving work quality; it’s about “gaming.” He’s using AI to optimize his communication with superiors, not the actual work.
2. AI-Rated 360-Degree Reviews: Employees use AI to write self-ratings, and colleagues use AI to write evaluations. The boss reads a bunch of nicely worded comments generated by AI.
- Essence: Performance assessments have become a formality. When everyone uses AI to generate perfect evaluations, they lose their value as genuine feedback and turn into a “word game.”
In simple terms: It’s like cheating on an exam. Before, it was copying answers; now, it’s having AI write the answers and another AI check them. People are no longer competing in their understanding of the business but in their ability to deceive AI and bosses. This kind of competition is meaningless and only consumes energy and the company’s Token resources.
---
5. The Ultimate Question: What Is the “Residual Value” of Humans?
Finally, the report raises a philosophical question: When execution becomes extremely cheap (AI can write code and reports in seconds), what’s the value of humans?
The article’s answer is clear: **Human value lies in “judgment,” “responsibility,” and “coordination.”
1. Defining problems: AI can answer “how to do it,” but only humans can answer “why to do it.”
- For example, a data analyst like Keke can use AI to write SQL queries, but AI doesn’t know which data sources are unreliable or what the business really needs. Humans define the problems, and AI solves them.
2. Taking responsibility: AI can make mistakes, but it doesn’t have to bear the consequences.
- If AI causes system failures or misleading reports, the person who gets fired or has their performance deducted is the human, not the AI. This ability to take risks is irreplaceable by AI.
3. Coordinating resources: AI can create perfect communication scripts, but it can’t get two conflicting departments to sit down and talk or get the boss to approve a budget. Humans act as “connectors,” while AI is an “accelerator.”
In simple terms: In the AI era, the value of “executors” has decreased, while the value of “decision-makers” and “responsible individuals” has increased.
- If you’re just an executor (coding, writing, creating tables), your value is diminished by AI.
- *If you’re a decision-maker, coordinator, or someone who can identify and correct AI’s errors, your value increases significantly.
In summary:
This AI revolution isn’t about eliminating workers; it’s about “selecting” them.
- Those who can only execute mechanically, refuse to think, and avoid responsibility will be eliminated.
- Those who can harness AI, define problems, and take responsibility will be retained.
Advice for everyone:
Don’t panic, and don’t blindly chase AI usage.
1. Learn to ask questions: Don’t treat AI just as a search engine; use it as an intern and learn to break down tasks.
2. Maintain critical thinking: Never fully trust AI’s outputs. Your core value lies in identifying errors and making decisions.
3. Emphasize responsibility: Take the initiative to handle the “gray areas” that AI can’t cover (e.g., cross-departmental coordination, crisis management). These are your strengths.
**AI is a tool, not the master. Don’t become its battery; be its commander.”