Summary of Key Points
Feng Bingqing, the CTO of Huajiao, shared the company's efforts in advancing "organizational AIization" (the use of internal agents). Although employees individually use AI tools such as ChatGPT and Claude, overall efficiency has not increased. The reason is that the use of AI remains at a "single-point, individual behavior" level. Huajiao is trying to transform AI from a personal tool into an integral part of organizational processes. During this process, they have encountered issues such as employee anxiety about being replaced, SOPs (Standard Operating Procedures) not being understood by AI, friction in departmental collaboration, and cost pressures. They have addressed these challenges through training frontline employees, modifying SOPs, integrating AI usage into performance evaluations, and controlling token costs. The ultimate goal is to make agents operate autonomously. Feng also predicts that we will enter an era where agents will be paid for their services, and AI will further widen the gap in information access.
Detailed Analysis
1. Why isn't overall efficiency improving despite everyone using AI?
Many companies have employees using AI tools, but when reviewing business performance, there is no noticeable increase in efficiency. Feng Bingqing believes the problem lies in the "single-point use" of AI: different people use AI in various ways and capacities, and these individual practices are not integrated into organizational processes. For example, technical staff use Codex to write code, while operations staff use ChatGPT to create content, but these activities remain isolated within departments and do not contribute to overall efficiency improvement. To enhance efficiency, AI must be transformed from a personal tool into an organizational one, integrating it into the company's business processes rather than just being used for personal purposes.
2. The biggest obstacle to organizational AIization is not technology, but employees' concerns
The main barrier to implementing AIization is not technical aspects (such as interface integration or permission management), but rather employees' perceptions and emotions:
- Non-technical staff: They do not understand how to use AI and fear being replaced by it (for instance, designers may require that AI-generated content be revised manually to prove their value).
- Technical staff: They are anxious about rapid technological changes and worry about being displaced.
- Middle management: They find themselves in a difficult position between managers who demand efficiency and employees who express concerns.
Huajiao's solutions include:
- Starting with frontline employees by showing them practical examples of how AI can improve work (e.g., handling urgent tasks without needing to use computers).
- Providing basic AI training and teaching employees how to use it in specific contexts.
- Clarifying that "AI freeing up manpower does not mean reducing staff"; instead, it aims to liberate employees from repetitive tasks (such as data retrieval or basic writing) so they can focus on more creative work (e.g., developing 3D animated gifts).
- Incorporating AI usage into performance evaluations to encourage employees to learn.
3. Converting human work processes into language that AI can understand
Each department in the company has SOPs, but these are either stored in employees' minds or written for human readers—AI cannot understand them directly. For example, employees know the steps required to retrieve information about streamers, but AI does not know how to access the relevant backend interfaces.
Huajiao's technical team has played a crucial role by translating these SOPs into commands that AI can execute. Programmers are no longer just writing code; they have become "business architects," transforming employees' experience into logic that machines can understand. For example, when retrieving streamer information, the technical team connects the backend interfaces to agents, allowing employees to simply communicate with the agents via DingTalk, and the agents automatically handle the data retrieval while controlling access rights.
4. The ultimate goal: creating a framework for autonomous AI operation
Huajiao's long-term vision is to create a self-sustaining system where agents can operate without human intervention:
- Cost control: Managing token consumption (the unit of payment for AI services), and switching to cheaper AI products when budget limits are reached.
- Departmental integration: Connecting departmental processes with agents in the first half of the year, and then enabling automatic coordination between agents from different departments (e.g., business and operations) in the second half of the year.
- Memory-free design: Huajiao's agents do not have a memory function because organizational processes are standardized; there is no need for AI to remember individual preferences. The focus is on establishing company-wide consistency (e.g., all requirement documents follow the same standards).
5. Two predictions about the future of AI
Feng Bingqing believes that:
- An era of agent-based payments: In the future, agents will function like applications, with each serving a specific purpose (e.g., an "agent for retrieving streamer information" or "an agent for writing content"). Companies will pay based on the number of times these services are used.
- Widening information disparities: AI will exacerbate the gap between people who use it extensively (e.g., those who can automate complex processes) and those who do not, leading to a hierarchical structure where not everyone can equally benefit from its advantages.
This analysis explains Huajiao's AI practices in simple terms and highlights the core issues and future trends of organizational AIization, making it accessible to non-financial professionals as well.