Hello! I'm your financial analysis assistant. This article from "Tsinghua Management Review" aims to bring a reality check to all HR professionals and managers who are struggling with the impact of AI, while also offering some practical solutions.
In simple terms, today's companies are busy acquiring and utilizing AI, but most of them haven't figured out how to effectively manage it. The core argument is that AI is no longer just a simple tool; it has become a "digital employee." If you continue to manage it using traditional methods, it will lead to chaos. The winners of the future won't be those that use the most AI, but those who first understand how to combine human skills with AI in a way that doesn't alienate their employees or overlook their responsibilities.
Let me break down this article into five key points in plain language:
1. The current situation is awkward: Companies want to save money, but employees are struggling
First, we need to face a harsh reality: the pace of AI adoption in companies is much faster than the learning rate of their employees.
Many companies are being very hypocritical:
- At the company level: They invest heavily in AI tools and deploy intelligent systems to reduce labor costs and increase efficiency.
- At the employee level: Employees are expected to use AI, and its use is even included in performance evaluations. However, companies often don't provide adequate training, permissions, or support.
This creates a significant structural contradiction:
- Who pays the cost? The company (for software, computing power, and training).
- Who reaps the benefits? The company (from reduced wages and increased efficiency).
- Who bears the pressure? The employees (from having to learn new skills, fearing replacement, and increased workloads).
Data is alarming: 42% of American employees feel that their companies expect them to learn AI on their own, and 81% feel more stressed at work. It's like the boss gives you a new machine and tells you to use it without teaching you how to operate it, threatening to fire you if you don't do it well. This approach shifts the risks of transformation onto the employees while keeping the benefits for themselves. This is not just a management oversight; it's an unfair design of the profit distribution.
2. The role of AI has changed: From a "calculator" to a "colleague"
Many managers still think of AI as a high-level calculator, but the reality is different. The article outlines four stages of AI integration in companies, and many are stuck between the second and third stages, yet they still use first-stage thinking, which is the root of the confusion:
- Stage 1: Tool (calculator). AI helps with tasks like writing emails and searching for information. At this point, you only need to manage how it's used; humans are still the main focus.
- Stage 2: Collaborative partner (intern). AI starts to participate in analysis, content generation, and decision-making. The question arises: Who is responsible for mistakes? Is it the AI's fault or the human's? You need to redefine the boundaries of responsibility between humans and AI.
- Stage 3: Digital employee (full-time employee). AI has specific roles, such as customer service or code testing. Now, AI is a labor force with defined responsibilities. Traditional job descriptions and performance evaluations no longer apply.
- Stage 4: Multi-agent systems (teams). Multiple AI systems work together, with each performing different tasks. This leads to complex collaboration among humans, humans and AI, and AI and AI. Management complexity increases exponentially.
The key conclusion: You can't manage a team using the same methods as you would a calculator. If your understanding is still at the "tool stage" while you're already deploying multi-agent systems, it will result in chaos.
3. How to manage AI? Don't assign roles; define capabilities
This is the most counterintuitive but also the most insightful part of the article. Many companies name AI systems and assign them roles, like "Product Manager AI" or "Test Engineer AI," and expect them to work like human teams.
This is a big mistake!
- Why? Humans need division of labor because our attention is limited and there are professional barriers. But AI (especially large models) doesn't have these constraints. One model can handle both coding and writing copy. Forcing AI into specific roles not only prevents it from achieving its full potential but also hinders its effectiveness.
- More serious consequences: Information transfer between AI systems is often limited to conclusions. If A provides a conclusion to B, and B passes it to C, the reasoning process is lost, leading to cumulative errors. Even if each step seems correct, the overall result may be wrong.
The correct approach is:
- Avoid role-based division; instead, focus on context isolation and tool modularization.
- Define the behavioral guidelines and characteristics of the AI system.
- Specify the tools it can use and the data it can access.
- Control precisely what information it can see and what resources it can operate on.
This is like managing a team: it's not about titles, but about what capabilities and access rights each member has.
4. Where has the value of humans gone? Don't cut out your experienced employees
As AI becomes more powerful, repetitive and rule-based tasks will be automated. Some companies drastically reduce their staff, thinking they're more efficient.
But this calculation is flawed:
- Hidden costs are ignored: You're not only losing employees but also their " Tacit Knowledge"—the unspoken expertise that's hard to quantify, such as medical expertise, negotiation skills, intuitive judgments in crisis situations, and long-term customer relationships.
- What can AI learn? AI can learn explicit processes and documents, but not the experience and intuition that even the creators can't articulate.
- The lesson from Klarna: This Swedish fintech company replaced most of its customer service with AI, only to find a decline in customer satisfaction. The CEO realized they went too far and reintroduced human services.
- Blind spots: Companies only consider the immediate cost savings, ignoring the operational costs of AI (computing power, maintenance), error handling, and the loss of business due to decreased customer satisfaction.
The future value of humans lies in:
1. Task definition: Clearly defining what AI should do and breaking down tasks, which requires a deep understanding of the business.
2. Result verification: Assessing the accuracy of AI's decisions and identifying risks, which requires experience and intuition.
3. Responsibility: In case of errors, only humans can be held accountable.
Therefore, don't dismiss experienced employees; they are the organization's resilience in the face of uncertainty.
5. Reconstructing the system: HR needs to become a "human-machine relationship coordinator"
The article proposes a new management framework: HR&MAM (Human Resources and Multi-Agent Management).
Traditional HR focuses on people; now, it needs to manage three aspects:
1. Human resource management: Reevaluate traditional approaches and focus on the transfer of value and the reconstruction of employees' skills.
2. Agent management: Assign responsibilities to AI, calculate costs, and manage risks. Each AI should have a cost-benefit analysis, just like employees.
3. Collaboration interface governance: This is a new area that involves managing the interaction, supervision, and ethical boundaries between humans and AI.
There's a significant ethical issue here: knowledge transfer. Companies often ask employees to share their expertise with AI, which some see as a self-destructive strategy (like "sawing the branch you're sitting on"). This raises three ethical questions:
1. Consent: Do employees truly understand and agree to have their knowledge replicated?
2. Ownership: Who owns the value created by AI? Does the AI's "digital clone" continue to generate revenue for the company after the employee leaves?
3. Legality: Using AI to retain employees' knowledge weakens their bargaining power.
Future leaders must:
- Be transparent: Clearly define data usage and inform employees.
- Be fair: Share the benefits of AI with employees (through bonuses, retraining, or better working conditions).
- Show respect: Recognize the value of tacit knowledge and protect employees' job security.
In summary:
In the AI era, technical implementation is just the beginning; organizational restructuring is crucial. If you focus only on efficiency and ignore fairness, responsibility, and human dignity, the savings you make will be lost due to trust erosion, talent loss, and management chaos. The truly successful companies are those that first balance humans, AI, and systems, always respecting the value of individuals.