Summary of Key Points
This article illustrates the pitfalls that individuals and organizations can easily encounter when using AI in the AI era through the story of a training and education business owner who failed to develop a subtitle tool using AI programming. Individuals tend to skip basic verification and jump straight into complex projects; boss-type product managers confuse the processes of coming up with ideas with defining clear requirements. Additionally, if organizations implement AI across the board without the necessary mechanisms for requirement breakdown and evaluation, they may experience a false sense of progress where individual efficiency increases, but the organization as a whole does not improve. The article also outlines the evolutionary path for organizations that are truly AI-native, emphasizing the need for organizations to have the ability to determine whether to use AI, how well it is being used, and when to stop using it in order to fully leverage its value.
1. AI Programming is Not a “Wand”: Three Common Misconceptions When Using AI Individually
The training and education business owner in the case initially thought that AI programming was incredibly powerful and quickly created an app framework. However, numerous problems arose later on, such as slow video processing, high consumption of tokens, and difficulty understanding the AI-generated code. Why? Because the owner made three typical mistakes:
- Rushing into large projects without small-scale verification: For example, instead of trying to use Whisper to process a 10-minute video to assess accuracy, time, and cost, the owner went straight to developing the complete app.
- Vague requirements: The owner wanted students to be able to watch English movies more easily but simply defined the requirement as “a multi-model subtitle app” without considering simpler solutions like using local scripts or web plugins.
- Lack of acceptance criteria: There were no clear standards for things like processing speed and accuracy, leaving the AI to keep adding code indefinitely, resulting in a project that became increasingly complex without resolving any issues.
2. The “AI Illusion” of Boss-Type Product Managers: Ideas Are Not Equivalent to Executable Requirements
Many bosses believe they are the company’s primary product managers and are dissatisfied with the products created by their employees. However, with the advent of AI, they realize that they often only know how to state their needs (e.g., “I want a subtitle tool”) without being able to articulate them in a way that AI can understand, such as “the tool should handle MP4 files, achieve an accuracy of over 90%, and cost no more than 5 yuan per movie.” In the past, these tasks were handled by product managers and developers, but now AI has pushed bosses directly into the front line, exposing their core issue: their inability to define clear requirements.
3. Implementing AI Across the Board Does Not Mean Everyone Using ChatGPT: Two Shortcomings That Organizations Need to Overcome
Many companies are advocating for the widespread use of AI, but the result is often that everyone ends up developing their own small tools, leading to chaotic cross-departmental collaboration and a proliferation of systems that are not effectively utilized. The problems lie in two areas:
- Lack of requirement breakdown skills: Employees who are close to the business process do not know how to break down tasks into steps that AI can execute (for example, the owner in the case could not analyze the technical feasibility or potential formats for the subtitle tool).
- Lack of evaluation and management mechanisms: No one determines whether these tools are necessary or effective, nor is there anyone to manage shared resources (such as a unified speech recognition service).
The correct approach is for frontline staff to identify issues and conduct small-scale verifications, while specialized teams develop complex systems, with managers setting goals and standards.
4. The Five Stages of Evolving into an AI-Native Organization
The article outlines five stages of developing an AI-native organization:
- Stage 1 (Individual Tool Use): Employees use AI to write content and code, improving individual efficiency without changing organizational processes.
- Stage 2 (Process Standardization): AI is integrated into stable processes (e.g., video upload → recognition → translation → export) with clear input and output standards for each step.
- Stage 3 (Team Collaboration): Clear divisions of labor among bosses, product managers, developers, and operators create a streamlined workflow (student feedback → requirement identification → development → testing → deployment).
- Stage 4 (Digital Employees): AI takes on continuous tasks (e.g., processing videos and generating vocabulary exercises).
- Stage 5 (Organizational Infrastructure): Establishment of unified resources (e.g., shared models and data standards) to avoid redundant development.
5. The True Value of AI for Organizations: Optimizing Processes, Not Just Reducing Staff
The purpose of AI is not to turn everyone into programmers; rather, it helps organizations with three key tasks:
- Automating repetitive tasks: Customer service inquiries, meeting minutes, and other repetitive tasks can be handled efficiently with the help of AI.
- Reducing Handoffs: In the past, requirements had to go through multiple departments, leading to information loss; AI can directly capture original requirements, reducing errors.
- Providing Evidence: AI can record task quality and response times, helping managers make informed decisions.
To determine if a company is truly AI-native, look for three indicators: whether AI is a stable part of the workflow, whether it has led to a reduction in the number of positions and handoffs, and whether there are unified evaluation mechanisms.
In summary, the evolution of organizations in the AI era revolves around enabling employees to use AI effectively and ensuring that organizations know how to manage its use. Otherwise, implementing AI across the board may result in a superficial increase in efficiency, with everyone busy with individual tools without real organizational improvement.