Summary of Key Points
This article reveals five “heartbreaking truths” about the AI industry: Many AI companies are merely surviving through custom projects (and are essentially dead in terms of operational efficiency); the key to AI transformation lies in company leaders personally adopting and utilizing AI; most of the data accumulated by enterprises is of poor quality; product managers who rely solely on AI to create prototypes cannot succeed; and AI-powered products without a clear business logic are doomed from the start. The core message is that AI does not challenge established business principles; rather, it exposes long-standing issues that were previously masked by funding and fancy presentations—issues such as the quagmire of custom projects, leaders’ lack of understanding, data-related problems, the absence of a product-oriented mindset, and flaws in business logic. These problems will not disappear simply because they are labeled as “AI-driven.”
Detailed Analysis
1. Custom Projects: The Quagmire of AI Companies’ Survival
Many AI companies seem to be doing well financially, but their success is largely due to custom projects. However, this creates a vicious cycle:
- Preliminary Sales: To secure contracts, companies agree to almost anything—clients demand intelligent systems, knowledge bases, automated reports, and integration with ERP/CRM/OA systems, often adhering to outdated business practices. The contracts are signed eagerly.
- Project Delivery: Once the contract is signed, it becomes clear that there are significant issues: model performance is unstable, client data is disorganized, business rules are unclear, and requirements change frequently. As a result, AI projects turn into sources of conflict—financially attractive but with delayed acceptance and payment.
- Survival Cycle: Before one project’s problems are resolved, another requires even more extensive preliminary efforts to fund the delivery of the previous project. This is not growth; it’s a mere act of delaying the inevitable demise.
2. The Risk of AI Transformation: Leaders Who Don’t Understand AI
The AI transformation in many companies merely involves implementing a few projects and using tools, but the crucial factor is that leaders must personally embrace AI:
- For example, in software companies, although design, development, and testing processes are being automated with AI, leaders may be reluctant to allocate funds for AI tools because they have never used them themselves and therefore doubt their effectiveness.
- Without a deep understanding of AI, leaders cannot distinguish between poor tools or employees’ lack of skills, or determine whether the issue lies with the models or the wrong choice of use cases. As a result, the direction of transformation is misplaced, rendering all employee efforts futile.
3. Corporate Data: Not a Gold Mine, but a Heap of Trash
Companies often believe that their decades-old PDFs, Excel files, and Word documents represent a valuable data asset, but half of this information is useless:
- The same field might be referred to as “number of customers” by sales and “number of orders associated with customers” by finance; the same process may be described differently in policies and actually executed differently. Important rules are hidden in notes or merged cells—human employees can guess or ask experienced staff, but AI cannot.
- Consequently, 80% of the time in many AI projects is spent organizing data: standardizing definitions, filling in missing information, and cleaning up discrepancies. This is not a reflection of AI’s limitations; rather, it’s the result of past data-related issues that now need to be addressed.
4. AI Cannot Save Product Managers Who Rely on Prototypes
Many product managers think that mastering AI for prototype creation and requirement writing will lead to success, but this is a misconception:
- While AI can improve efficiency, without a product-oriented mindset, such efforts are futile. For instance, instead of creating pages based on leadership instructions, the task now involves feeding those instructions into AI tools. In the AI era, product managers must understand why customers pay, how business processes work, and what real problems users need to solve; otherwise, even with advanced AI, their incompetence will be exposed.
5. The Failure of AI-powered Products
Many AI products are doomed from the start because they fail to address fundamental business issues:
- Overwritten by Larger Models: Some products may be temporarily popular due to the absence of certain features in larger models, but once those models evolve, these products become obsolete (e.g., AI writing tools could be replaced by more advanced systems like GPT).
- Lack of User Value: Beautiful demos and impressive launches do not address the core question: why should customers pay? Why continue paying? Why not use free alternatives?
- Lack of Competitive Advantages: When competitors lower prices, these products are forced to follow suit; when customers compare options, they find little difference—ultimately, they are defeated by price wars.
AI has changed the technical landscape but not the fundamental nature of business. Products without user value, a viable business model, or competitive advantages will not survive, even with an AI label.
In Conclusion
AI is not a magic solution. Companies that rely on custom projects to stay afloat will continue to struggle, and those product managers without a solid understanding of business will remain ineffective. The underlying problems will not disappear just because they are labeled as “AI-driven.”