Why Employees Resist AI at Work: Lessons from a Podcast with Zhou Hongyi

1. Introduction

On September 24, Luo Yonghao released episode 5 of his podcast "Luo Yonghao's Crossroads", featuring a conversation with 360 founder Zhou Hongyi.

Zhou said the company is building a mindset for employees to use AI, insisting on its use even if results are not great.

The company runs internal AI competitions; there have been no mass layoffs due to AI, but employees who refuse to use AI after it is mandated may be let go.

2. Why Do People Reject AI?

People’s feelings toward AI vary: some find it extremely helpful and study it intensely, while others see no value and resist using it.

The author offers personal understanding.

2.1 Technological Development Stage Limitations

Analogy of progress from horse‑drawn carriages to rockets, from BB machines to smartphones, from machine language to natural‑language interaction.

Early technologies may be slower than older ones (e.g., cars initially slower than horse‑drawn carriages).

Many AI products are hyped by media; negative cases are rarely highlighted, leading to disappointment when real performance falls short.

Current large models are stronger, but many claimed "digital employees" or "intelligent agents" still cannot match even basic tools like Copilot, which reduces acceptance.

2.2 Lack of Best Practices

People interact with AI via natural language, mistakenly assuming AI can work like a human.

Models can hallucinate, performance degrades with long context, and results depend heavily on the user's expression ability and experience.

Example: tasks that AI could handle well fail when users do not provide clear instructions or sufficient context, leading them to think AI is ineffective.

Many AI products focus on rapid development and flashy demos, neglecting to provide users with best‑practice guidance, making onboarding difficult.

2.3 “Learning” vs. Human Nature

Quote: "Work with human nature, grow against it."

Human nature prefers laziness, praise, and strength; learning is hard, especially after entering the workforce.

2.4 Many Have Not Been Touched

The author often uses a diagram in company talks to illustrate that most people would choose a round wheel if they are not "stupid".

However, at the current stage many still behave that way.

Examples: DeepSeek has higher hallucination rates; people avoid trying Claude, Gemini because they cost money.

Many only ask simple questions in the input box; when suggested to build an intelligent agent for complex tasks, they reply "this is enough for me".

This is my simple explanation of an expert agent that learns knowledge extremely fast.

This is my article‑to‑card assistant that makes reading articles very convenient.

These examples are numerous. When the author shares AI experiences and trains colleagues, many report huge benefits and excitement, making it hard to imagine anyone refusing AI.

Once people truly experience AI’s usefulness, resistance fades and they may even lead AI adoption.

2.5 Learning Mode Problems

In exam‑oriented education, scores dominate; books unrelated to scores are "junk", and unrelated activities are "waste of time". This mindset is deeply ingrained.

After entering the workforce, KPI becomes king; books unrelated to KPI are "unproductive". Companies promote learning AI, but many employees’ KPI/OKR do not align with AI usage, resulting in limited impact.

3. Conclusion

The early AI era feels magical. Some still work with old methods, while others become "super individuals".

Some embrace AI fearing obsolescence; others dismiss it as "nothing".

Viewing AI development with a long‑term perspective, model capabilities will keep improving, offering greater assistance.

In the short term, those who actively use AI and master best practices will gain the biggest advantage.

What do you think about this?