Hello! I'm your financial analysis assistant. Although the title of this article contains technical terms like "FDE" and "ontology," which might sound quite complex, the core message it conveys is a topic that every boss, HR professional, and manager should be concerned about: How should companies adapt when AI truly starts to take on tasks?
Many bosses currently have a misconception that just by purchasing AI tools and training employees to use them, the company has successfully transformed. However, the author, Wei Haozheng, points out that this is like equipping a horse-drawn carriage with an engine, but keeping the wheels made of wood—meaning the company won't be able to move quickly and may even fall apart.
To help you understand this transformation completely, I'll break down the article into five key aspects in plain language:
---
1. Why hasn't the efficiency of your company improved despite having AI?
Core point: Individuals have become faster, but the organization hasn't kept up, leading to increased internal friction.
Many companies find that employees are indeed using AI to complete tasks more quickly, such as writing code or creating PPTs. However, bosses may not notice much change, or things might even become more chaotic.
- Phenomenon: Instead of one person handling one task, one person now uses AI to handle multiple tasks. But the old departmental boundaries and approval processes remain in place.
- Contradiction: Department A uses AI to process data rapidly, but Department B still approves things slowly according to the old rules. The data from Department A might not be received by Department B, or it might not be passed on due to permission issues.
- Conclusion: It's like giving an old-fashioned carriage a rocket engine. The "horse" (employees) can run fast, but the "axles" (the organizational structure) are still outdated. As a result, either the carriage breaks or internal friction increases due to the high speed.
- Formula: Organizational AI productivity = Individual AI capability × Organizational AI adaptation. If the adaptation is zero, no matter how strong the individual capabilities are, the overall output will be zero or even negative.
2. What is FDE? Not a senior programmer, but a "resident doctor"
Core point: FDE (Frontier Deployment Engineers) act as translators and problem solvers between technology and business.
Many people think FDE refers to more advanced programmers. However, this is not the case.
- Traditional approach: Customers submit requirements → Product managers write documents → Development teams write code → Testing → Deployment. This process is slow, and the final product often doesn't meet customer needs.
- FDE approach: FDE teams go directly to the company's site. They don't rush to write code; instead, they first:
1. Understand what the business is doing.
2. Identify the biggest challenges (e.g., why does the approval process involve seven people? Why don't the data match?)
3. Determine which steps can be automated with AI.
4. Plan how to integrate AI.
- Value: FDE teams are small, consisting of people with technical, business, and organizational expertise. They diagnose the problem, propose solutions, and quickly develop prototypes.
- Change: A project that used to take 8-12 people months to complete now can be done by a FDE team in just a few days with AI assistance. This is not just a technical upgrade but also a revolution in the way work is delivered.
3. What is "ontology"? Giving AI "expert eyes"
Core point: Ontology is not some mysterious technology; it's about turning the company's hidden knowledge into a structure that AI can understand.
This is the concept that's often misunderstood in this context. Ontology sounds philosophical, but in the context of AI, it's very practical.
- Challenge: AI is smart, but it doesn't know how your company operates. It knows the word "order," but it doesn't understand that orders over a certain amount must be approved by the finance director, or that customers are categorized differently with varying discount rights.
- Role of ontology: Ontology organizes the business logic, relationships, rules, and permissions hidden in old employees' memories, Excel spreadsheets, and documents into a structured system.
- Objects: What is a customer? What is a product?
- Relationships: A customer buys a product, resulting in an order.
- Events: An order over a certain amount triggers an approval process.
- Permissions: Only the director can approve.
- Essence: Creating an ontology is a process of self-examination for the company. Many bosses can't clearly define who is responsible for what or under what circumstances exceptions can be made. AI forces them to clarify these ambiguous points.
- Key phrase: Without ontology, AI only knows many facts but doesn't understand the context of the business.
4. Rebuilding the organizational structure: From a pyramid to a network of capabilities
Core point: AI makes hierarchical structures obsolete. Future organizations will be dynamically composed around tasks and capabilities.
Traditional companies have a pyramid structure: people → positions → departments → levels. This is because people's capabilities are limited, and information needs to be passed up through layers, making management more complex.
- Changes brought by AI:
1. Reduced communication costs: In a 3,000-person company, meetings and reports were necessary because everyone had different information. With an AI-based system, people can directly ask AI, "What are this customer's past orders?" "Who is holding up the approval?" This greatly reduces communication costs.
2. Blurred job boundaries: If AI can handle 70% of a job, what's the purpose of the remaining 30% of the position?
3. Smaller teams, greater efficiency: The more AI is used, the fewer people are needed, but the output increases. One person plus AI can form a small, efficient team.
- New organizational form:
- Positions → Capability nodes: Instead of being labeled as "sales" or "customer service," people are defined by their capabilities (e.g., customer insight, solution generation).
- Departments → Capability networks: Departments are no longer fixed; they are dynamic networks of people and AI agents called upon based on task needs.
- Managers → Coordinators: Managers focus on scheduling resources, tasks, and AI capabilities.
5. The ultimate question: What do people do in the AI era?
Core point: AI liberates not just jobs but the value of people. Don't view AI as a tool for cutting staff; use it to unleash creativity.
This is the most important point in the article.
- Misconception to avoid: Many bosses talk about AI first and ask, "How many people can we lay off?" This is shortsighted. If you only aim to save money, you'll lose the time, experience, and creativity that AI can free up.
- Example: A company reduced its after-sales team from 100 to 20 people but didn't lay off the remaining 80. Instead, they retrained them for pre-sales roles. These employees, along with AI, became more effective in sales and customer service.
- New roles for people:
- AI handles: Repetitive, scalable tasks (data organization, report generation, standard processes).
- People handle: Decision-making, creativity, building relationships, making value-driven choices, and taking responsibility.
- Irreplaceable responsibility: AI can provide solutions, but it can't take responsibility for mistakes or compensate for losses. Humans must have the authority and responsibility in the AI era.
- Redefining work: If an employee used to spend 200 hours on a task, and now AI can do it in 1 hour, the remaining 199 hours should be used to design more meaningful and strategic work.
---
Summary: The correct path for AI transformation
The author outlines a clear sequence for organizational restructuring:
1. Understand the business (Bring in FDE to identify challenges).
2. Structure the company (Create an ontology to clarify logic and permissions).
3. Reallocate work (Determine what to automate with AI and what to do manually).
4. Rebuild the organization (Break down departmental barriers and form capability networks).
5. Liberate human value (Let people do meaningful work and take responsibility).
In one sentence: AI transformation is not about buying tools; it's about changing the organization. First, change your mindset (understand the business), then the processes (create an ontology), and finally, the structure (reorganize the organization). Otherwise, you're just labeling your old system with AI, and the old system will still limit its efficiency.