This article is not a real guide, but understanding the upper and lower bounds of this generation's AI capabilities can help corporate drones avoid being phased out🤖💼.
This article is indeed a guide, especially if you forward it to your boss👨‍💼.

It's something we've likely all come to agree on: this generation's AI technology, as a productivity tool, has reached a level of capability that allows it to replace many human tools.

Now, there are two problems laid out on the bosses' desks:

  1. How to reduce costs (aka how to lay off employees💸🔪).
  2. How to increase efficiency (aka how to squeeze more out of employees with AI⏱️💼).

After six months of research, the capabilities of this generation's AI have been clarified (in terms of magnitude). So, today, I'll share our real-world experience regarding the survival state of various positions in the internet software industry in the era of this generation's AI.

👨‍💻 Programmers and Engineers

Programmers cannot be directly replaced

If you've seen articles or videos like "GPT-4 helped me, a non-coder, write a complete game," don't let them lead you astray.

First, it's important to clarify for bosses who are not familiar with various engineering technical issues that the most troublesome part of a programmer's job is not writing code, but figuring out how to solve a myriad of complex problems. These problems are as complex as the real business needs. Obviously, even the best GPT-4 cannot accurately understand the real-world complex needs, let alone write the code for it.

Currently, the strongest GPT-4 in terms of coding cannot directly perform the following tasks of a programmer:

  1. Designing real, functioning software architecture
  2. Writing engineering-level code
  3. Understanding business requirements

However, the current GPT-4 can assist programmers in the following ways:

  1. Helping programmers design software architecture
  2. Assisting programmers with function-level code writing

Take the second point, for example:

Sometimes, I encounter some problems that are not part of the regular business code implementation. They are not difficult, but I am not very familiar with them. If it were before, I might need to study a bit and spend about half an hour. But with the help of ChatGPT, it can be done in five minutes.

Now, there's a lot of coding that's tedious but not mentally demanding, like the common model definitions in CRUD programs. Previously, doing it myself would take about half an hour. But with the help of Github Copilot, pressing a bunch of Tab keys can finish it in five minutes.

In short, this generation's AI is a very qualified assistant for programmers, with significant efficiency improvements. I recommend that bosses, if not concerned about information security, should include ChatGPT and Github Copilot in their programmers' toolkit.

However, programmers may be indirectly replaced

If the efficiency improvements mean that what used to take 3 days can now be done in 2, or what took a 10-person team can be managed by 5 within the deadline. If business needs do not change while efficiency increases, it will inevitably affect some people's employment.

👨‍🎨 Designers

High-level designers cannot be directly replaced

Similar to programmers, the current AI cannot complete "engineering-level" design work. For example, brand design, product UI design, which requires maintaining a consistent overall design quality, is beyond the current capabilities of AI.

Moreover, the current AI's understanding of prompts in the design and art creation field is somewhat high, which might be challenging for the average boss. Don't think you can manage on your own; professional domain understanding by designers is still needed to effectively utilize these AI tools.

For instance, without knowing which design style belongs to which designer, it's impossible to use the designer's name to describe that style. Furthermore, if you know a designer but they haven't worked on a specific theme, then AI is not very good at creating in that theme (currently).

In summary, by using AI, designers can produce various schemes more quickly and in greater volume for evaluation, significantly improving efficiency. I recommend that bosses, if not concerned about information security, should include tools like Midjourney and Stable Diffusion in their designers' toolkit.

Tool-level designers will be replaced

Specifically, illustrators, original game artists, etc. The current AI, such as Midjourney and Stable Diffusion, outperforms humans in terms of speed and quality of output, making it feasible to downsize these positions.

Designers referred to as "graphic designers" within some teams are more likely to be replaced.

In short, if you're in this profession, consider your future as soon as possible.

📈 Product Planning and Operations

**

Senior product managers cannot be directly replaced**

Similar to programmers, the current AI cannot complete "engineering-level" strategic planning, nor does it have the capability to perceive the subtleties of business. However, AI can greatly improve the efficiency of product managers.

In the following areas, AI is stronger than a product assistant: in the speed and understanding of data analysis; in the speed and creativity of copywriting (creativity limited to English); in researching a field's support and assistance.

Using AI can improve their work efficiency. I recommend that bosses, if not concerned about information security, should include various AI tools in their product managers' toolkit.

Less talented product assistants will be directly replaced

As mentioned earlier, AI outperforms product assistants in many areas. If product assistants do not recognize this and consider their own value, then product managers would rather command AI than them. So, please consider your future as soon as possible.

🤝 Business, Marketing, Sales

These three positions in software companies are oriented towards people rather than machines, often requiring offline work. All offline work cannot be completed through AI, so their replaceability is generally low.

However, the following situation is an exception:

For many small companies, channel development is crucial. Often, the biggest channels are the boss's relationships, and employees only support work on existing channels without contributing much to new channel development.

In such cases, replacement is possible. Because these employees, although interacting with people, essentially exchange their time for the boss's time to maintain channels. With AI, the time cost of maintaining channels is reduced. Previously, one person could maintain 5 channels, but now one person can handle 20. Therefore, without expanding business scale, it's possible to cut 3/4 of the staff.

💀 Remote Customer Service

You might say customer service is about interacting with people... so... haha, I know what you're thinking.

The replacement of customer service has already begun a few years ago. Only this generation's AI has significantly reduced the cost of replacing customer service and improved performance.

Except for a few customer service specialists needed to maintain the knowledge base, the value of those "operator" type customer service roles has diminished to zero. Don't hesitate, please consider a career change :)

By the way, this generation of AI actually delivers a blow to companies that run traditional AI customer service businesses. Since all companies can easily build their own customer service robots, the technical and price advantages of many traditional AI customer service companies no longer exist.

🤵‍♀️Management Personnel

Management personnel cannot be replaced by AI, but there are many reasons to downsize them, not necessarily because of AI

The value provided by different management personnel varies. Some bring business or people, some can integrate and coordinate resources, and some are just figureheads but popular.

None of these can be addressed by this generation's AI. But good managers can still improve their work performance through AI.

👨‍💼 Bosses

The impact of AI on bosses is significant and requires special attention:

  1. Bosses might realize artificial intelligence is a trend and start deploying and directing personally.
  2. Bosses might not fully understand what this generation of artificial intelligence can and cannot do.
  3. If a boss's business is disrupted by artificial intelligence, their company might close down.
  4. If a boss's company closes down, then all positions in the company will be downsized.

So, for your own happiness, please forward this article to your boss. Cheers.

On a Positive Note

Actually, the strong supportive capabilities of this generation's AI will give rise to more One man armies. A specific example can be seen in the gaming industry.

In 2020, a thrilling game called "Project: Wingman" was released on Steam to rave reviews. The game excels in all aspects, with the key being that its entire team consisted of only three people.

Over the years, various engines have made the technical difficulty of game development relatively low, allowing developers to focus on making games more fun. Of course, if your dream is to develop engines, then there are relatively fewer positions available than before, as every game developing its own engine was a trend of the last century.

Now, other industries are experiencing what the game industry did back then; some professions will disappear, but many will be reshaped.

Take designers, for example. Although it may be frustrating that years of hard work in drawing are outperformed by AI, don't forget that the essence of being a designer is to express inner emotions through suitable forms.

Drawing is one form of expression, and writing prompts is another. To bosses and audiences, there's no difference between the two forms of expression. The content of the expression is what's key, not the method of expression.