AI under classical product mindset: Are we AI-ready?


We aren't quite ready for AI, at least not in the context of "classic products" for most Internet companies.

Drawing from my years of experience in internet giants, I will share my insights and observations on this topic, based on real-world projects, cases, and lessons learned.


Because they are still adhering to the production model of classical product mindset.

Classical product mindset typically exhibits the following traits:

  1. Prioritizing the increase in user numbers through enhancing user experience, particularly favoring the use of small yet impactful features to achieve significant outcomes.
  2. Focusing on certainty of products/business.The deliverables of requirements always manifest as a specific feature, a defined product, underpinned by a clear logic.
  3. Prioritizing data. Identifying weaknesses from data analysis, and initiating the next iteration to continually enhance products/business quality.
  4. Valuing user feedback. Adeptly addressing user feedbacks by optimizing business logic and updating features. This expertise contributes to refining the user experience.

For AI products, they often exhibit the following characteristics:

1. Uncertainty

AI products often exhibit uncertainty in their outputs. For example, an AI model that can correctly identify tomatoes in most cases may still misclassify them as apples from certain tricky angles.

2. Continuous Optimization and Iteration

Unlike traditional products, AI products require continuous optimization and iteration to improve their performance. This process is not as straightforward as documenting changes in each feature for traditional products, but rather requires a long-term investment of human effort for continuous improvement. Progress may be incremental, making it challenging to articulate precise changelogs in AI products.

3. Long-term Patience and Vision

The iteration of AI products takes time and patience. In reality, there is often an overestimation of both personal and company patience. Various uncontrollable factors tend to push projects towards quick and easy solutions. However, for AI products, short-term and quick solutions often lead to project failure.


This is somewhat akin to classical physics evolving into quantum physics, where all classical mechanics laws become obsolete, and it feels like God is truly rolling the dice, with probability and uncertainty pervasive everywhere.

However, due to the immense popularity of AI, whether it's to stay competitive with peers, make the business more appealing, or simply to have a new story for the boss's report, there's always a reason to align your business with AI.

As a result, there's a dual sentiment towards AI. On one hand, there's a fascination with the trend, fearing the inability to catch up with the AI bandwagon. On the other hand, there's a concern that AI is ridden with uncertainty and might be unpredictable. In the short term, expectations for AI are often unrealistically high, as if AI can solve every business problem. In the long run, there's a lack of confidence, and even when AI products are developed, people often find themselves unsure how to effectively utilize them. Consequently, AI business and AI products become a tangled web of confusion and underutilization.

How to evolve from the classical mindset to a quantum mindset?

Here are some overarching strategies to consider:

1. Anticipate industry trends.

In a rapidly evolving field where long-term quantitative predictions are challenging, experienced tech professionals can provide relatively accurate short-term judgments. For instance, in order to assess the positive feedback or adoption rates for LLM question-answering in the short term, we can predict the improvement after prompt optimization, project additional enhancements after RAG optimization applied, or LLM model upgraded in the next half year. These are all judgable and quantifiable aspects.

2. Set realistic objectives.

Ideally, product managers should possess technical knowledge or at least a basic understanding of key technological principles. This ensures that they set reasonable and achievable goals that are grounded in reality, avoiding unrealistic expectations.

3. Mitigating Uncertainty and Risks

AI projects are inherently more prone to uncertainty and risks than traditional projects. Therefore, the critical question lies in how the project team and company respond when risks emerge. This serves as a test for the project management process itself and the overall business acumen.

4. Cultivating Patience and Vision

Both patience and vision are crucial for success. Lacking either can lead to failure. Patience without vision leads to stubbornness and inflexibility, while vision without patience results in missed opportunities and watching competitors succeed while reminiscing about "the good old days."

5. Data-Driven Decision Making

Scientifically established data metrics are essential for validating the correctness of strategic decisions and continuously guiding the project's direction. This is likely the most straightforward aspect for classical product managers.

Closing Thoughts

This article provides a high-level overview of the topic. In reality, there are many details to be discussed, considered, and addressed when applying classical product mindset to AI. I hope to have the opportunity to continue this series in the future and engage in further discussions with everyone on the topic of AI from a classical product perspective.

Leave a comment