With the rapid development of artificial intelligence technology, the emerging profession of AI product manager is gradually becoming a hot focus in the market. With the popularity of AI applications such as ChatGPT, many people are beginning to worry about the employment impact that AI may bring, however, Andrew Ng, the founder of DeepLearning.AI, sees a new opportunity – AI product managers will become one of the most promising and valuable careers in the future. This article will delve into the employment prospects, job profiles, core competencies, and market demand of AI product managers, helping readers fully understand the current status and future trends of this emerging profession.
As we all know, after ChatGPT became popular all over the world, the first reaction was: “Oh no, will my job be killed by AI?” “But there is one person who stands from a different angle -DeepLearning.AI The founder of Coursera, co-founder of Coursera, and a teacher at Stanford University, saw another direction:AI is not just here to grab jobs, it is also quietly opening a new golden career track.
In a recent open letter, he made it clear:AI product managers will be one of the hottest roles in the future. He said it bluntly: software is getting cheaper and faster, especially the cost of prototype development has been greatly reduced. Next, what the market needs most is not “how to do it”, but “what to do” is the most valuable——This is the core value of AI product managers.
Times have changed, and people who decide “what to do” have become scarce
In the context we are familiar with, Andrew Ng also gave a very vivid example: Ford knocked down the price of cars that year, and as a result, everyone’s demand for gasoline soared. There is actually an economic principle behind this phenomenon: when one “complementary” becomes cheaper, the demand for the other will increase.
The same applies to the current software industry:AI is like making the “car” of software development cheaper and faster。 Taking GitHub Copilot as an example, this tool can give you suggestions while writing code, which can directly improve the efficiency of programmers by writing code by 30%-50%. But this is only a superficial change.
The real change is that the entire development process has been rewritten. In the past, a product had to go through a complete set of “slow work and careful work” process from idea to launch. But now, with various AI tools, technological implementation is no longer a bottleneck.The biggest problem now becomes: Is this thing you want to make really valuable?
After 10 years of interaction design, why did I transfer to product manager?
After the real job transfer, I found that many jobs were still beyond my imagination. The work of a product manager is indeed more complicated. Theoretically, the work of a product manager includes all aspects of the product, from market research, user research, data analysis…
View details >
In other words, AI is not about putting “coders” out of work, but about making people who “think clearly about what to do” more important than ever –AI product managers do just that.
At present, through research, the industry has depicted three types of AI product managers:
1. AI Native Product Manager (Native AI Product PM)
Role positioning: This is the PM of the “frontal battlefield”, directly creating products with AI as the core, AI is the main body of the product, such as Midjourney, ChatGPT, Kimi, etc.
Core Competencies (User Defined): Super user insight ability, understanding prompts and AI capability boundaries, able to make decisions in a “semi-certain” state, and conduct rapid trial and error and iteration.
Research Supplement: These product managers are at the forefront of AI application innovation. Industry trends show that the evolution of AI agents and the popularization of multimodal large models will give birth to more AI-native applications, which are the main battlefields for AI Native product managers. Robin Li, founder of Baidu, believes that AI Native product managers need to have strong learning abilities, and may not all have a computer science background, but can quickly build prototypes for testing. The theme of the 2025 Global Product Manager Conference also includes generative AI and AI agents, directly confirming the need for such PMs.
Market correlation: For startups and tech giants committed to pushing the boundaries of AI-driven product experiences, AI Native product managers are an indispensable core force.
2. AI Infrastructure PM
Role positioning: They are the “behind-the-scenes heroes” and their main responsibility is to build underlying tools and platforms for technical teams such as algorithms and architectures, such as model training and deployment platforms, model monitoring and observability systems, prompt management platforms, and knowledge base services.
Core Competencies (User Defined): Familiar with AI/ML infrastructure principles, able to connect with technical teams and understand their development pain points, emphasizing system stability, scalability and technical docking capabilities.
Research Supplement: The work of this type of product manager is the basis for subsequent AI development and application. Some job postings describe roles focused on MLOps, distributed training pipelines, and inference infrastructure, which aligns well with the responsibilities of a platform-based AI product manager. For example, the role of a product manager at Google Cloud’s Security Graph, while focused on security, also reflects the platform-based responsibility of providing a core data/analytics platform. Alibaba’s “AI Clouder Program” also features “AI Infra” as a key direction, hinting at the market demand for product managers in this field.
Market correlation: For enterprises that invest heavily in self-developed AI capabilities or provide AI services in a platform-as-a-service (PaaS)/model-as-a-service (MaaS) model, platform-based AI product managers are crucial, as demonstrated by Alibaba Cloud Computing.
3. AI+ Product Manager (AI upgraded version of traditional products)
Role positioning: Most of these PMs embed AI as an “efficiency improvement tool” into existing product systems in traditional industries, such as intelligent customer service, search recommendation optimization, intelligent quality inspection, intelligent risk control, and internal process automation (RPA + AI).
Core Competencies (User Defined): Clarify business goals, determine whether AI is truly “improving efficiency”, collaborate with algorithm teams to evaluate implementation effects (such as accuracy and recall), and understand how to embed AI into existing product processes without disrupting the original logic.
Research Supplement: This role focuses on the practical applications and integration of AI. China Construction Bank recruits artificial intelligence application experts in retail business, responsible for promoting the innovative application and capability upgrade of AI in retail business, which is a typical case of AI+ product managers. NetEase’s recruitment of AI product managers to achieve the deep integration and application of AI in music, games, education and other industries, as well as the compound application talents in the fields of “AI + medical care” and “AI + manufacturing” mentioned in the Global Times report, all belong to this category. Alibaba’s “AI+” Career Trend Report also describes in detail how AI has penetrated into all walks of life, such as creating “smart breeding experts” through DingTalk and Tongyi Qianwen, which is a vivid embodiment of the AI+ model.
Market correlation: As industries accelerate their digital transformation and seek to leverage AI to gain a competitive advantage, AI+ product managers are in high demand.
Platform AI product managers, AI+ product managers are in high demand in the job market, enterprises are facing transformation, internal need to open source and reduce expenditure, and AI Native is the so-called AI native products, the current market is not much, there are few options for job hunting, and there are only a handful of people who can be named
Therefore, AI is no longer just the exclusive of large factories, and more and more traditional industries that originally have business, traffic, and foundation have also begun to enter the game, and AI+ product managers are becoming a key role in promoting the transformation of traditional enterprises. This shows one thing: AI is changing from a “black technology” to a “production tool” that everyone can use.
Although the position of “AI product manager” seems to be a bestseller, there are still structural problems in the market – Maimai’s data also mentioned that now in the AIGC industry, there are more people with non-technical backgrounds such as product managers who come out to find jobs, but what enterprises really lack is those who have the skills to work, which leads to a “mismatch between supply and demand”
For example, a set of commercially landable posters is actually a Comfyui workflow, a video face restoration is also a set of workflows, and a layer of UI interface is packaged outside, which becomes a usable product, in other words, because you have a sense of beauty, you understand AI, and you will complete the workflow after coming to the company , mass production is fine. So does this really need a person in a professional field to do it? Therefore, the vast majority of design positions, business positions, operation positions, product positions, trainer positions, etc., after joining the company, the work content is the same
Agent workflow + model training + prompt construction
For enterprises, there are too few people who understand the AI industry, have aesthetics, and can understand all the mainstream AI tools on the market, and collaborate effectively with the team, which may be no less difficult than finding a pure AI researcher.
So the following are the salary range, job search essentials, and ability requirements of the current companies I surveyed for AI product manager recruitment
Salary
When it comes to salary, this is probably the most important concern for many people. Let’s look at a set of data:
AI product managers, in the entire industry, the salary is not low. From January to August 2023, the average annual salary of domestic AIGC-related positions is more than 410,000 yuan, while those who specialize in AIGC product managers can get an average of 436,500 yuan. Another Morgan McKinley salary report is even more aggressive, saying that experienced AI product managers can earn an annual salary of 650,000 yuan. Of course, such a number is more senior, but it also shows that the ceiling of this position has been raised.
If we compare the salary level of the entire AI industry horizontally, the average salary of the AI industry is not low.
According to data from Zhaopin’s recruitment in the second quarter of 2024, the average monthly salary for artificial intelligence-related positions is more than 10,003, AI engineers can reach about 20,002, and in cities like Fuzhou, the monthly salary of AI engineers can even reach 20,004. **Although these numbers come from different channels and different job statistics (such as AIGC vs general AI, product manager vs engineer), one general trend is clear:In the AI industry, especially AI product managers who can “string” technology and business, the salary level is really not low.
Another particularly noteworthy is the so-called “AIGC premium”. Also in the direction of AI, the average annual salary of AIGC product managers is 436,500, which is even higher than the average salary of the AIGC industry as a whole (410,000).
According to the “2024 Chinese Intelligent Job Recruitment Research Report”, among all AI-related positions, candidates with 3-5 years of experience are the most vigorous, with a salary range of about 350,000 yuan, accounting for 31.67%. **At the same time, the competition is also the most fierce, the absolute majority are product managers who have moved over the industry through ToB or XR, VR, etc., they hope to be able to mix a job through past experience, in this case, if it is not the same vertical product line, candidates who need to have certain industry awareness and resources and project management capabilities are likely to have little employment opportunities.
So I compressed the direction of my research to 0-3 years of work experience (including career change), and my salary was controlled within the range of 250,000 annual salary. Smooth transition, quantify the plan of entry and exit
Now that I am an AI product manager, it is not enough to say that I know a little AI.You have to come up with some “real guys” – that is, a project portfolio that shows that you understand AI and know how to use AI.
The portfolio is not a show, it should make people see that you have really put in a lot of effort, such as:
Have you systematically learned what AI is all about? Have you thought about your career direction? Which technologies are still poor, should they be supplemented? Have you ever done some practical projects, whether it is done at work, done by yourself, or coursework while studying, as long as you can show your ability, it counts.
If you are switching from a traditional product manager, the focus is: how to combine your original product experience (such as research, analysis, and project promotion) with new knowledge in the field of AI. **Interviewers often look at whether you “understand products and AI”. Things like AI ethics, how to launch models, and how to continuously improve are also test points.
Now whether you have just graduated or transitioned from another position,Without a portfolio that “understands your AI strength”, it is difficult to enter the AI product manager track.
Just say that you have learned some theory, human-computer interaction, and recommendation system, and the company does not believe it. They want to see if you can really use this knowledge in projects to solve practical problems.
Graduates should pay attention to:Graduation projects and internship projects are best related to AI, and can be done;
Friends who are transforming, don’t just wait for the opportunity,Sometimes you have to take the initiative to find some AI-related projects to do, or set up a demo to practice your hands.
Another point, especially for those returnees or from foreign company backgrounds:If you really want to be an AI product manager in China, take more time to understand this set of “clay tools” in China.
For example, domestic frameworks such as PaddlePaddle and MindSpore, how to use Alibaba Cloud and Tencent Cloud, as well as collaboration software commonly used by enterprises.
To put it bluntly, even if you have seen the world and have a wide vision, you still have to be grounded when landing projects.Those who understand both general knowledge of AI and understand China’s technology ecology are more likely to be recognized by companies.
Because I am still testing this position, I need a lot of information to integrate. It involves data and privacy. So first simply show a page of portfolios.
Research directions
Integrate existing recruitment platforms for AI product managers
1) Understand the underlying principles of AI, not seeking proficiency but “clear door”
If you want to be an AI product manager, the first hurdle is to understand what AI is all about. You don’t need to be able to write advanced algorithms, but you need to know how machine learning and deep learning train models, why large models are so popular, and can deeply integrate them with business scenarios to explore innovative applications, token mechanisms, vector embeddings, and context length limits. It’s like a “common language” between you and the tech team.
For example, how is the model trained? How to go online after training? How to monitor whether it is “overturned” when it is online? You need to understand these basic processes so that you can judge whether a function is reliable, technically feasible, or a pipe dream. To put it bluntly, you don’t need to be able to write code, but you need to know the logic behind the code, otherwise you will “overturn” when you talk to the technical team.
Many companies now clearly write when recruiting AI product managers: you must understand the basic principles of machine learning, deep learning, and large models. As written in China Merchants Jinke’s JD, basic skills are the stepping stone.
There are many branches of AI technology, and the field in which you make products determines which technology you should focus on understanding.
For example, if you are a chatbot or intelligent customer service, you need to understand natural language processing (NLP): how to analyze text, how to generate content, and how to understand user semantics. What you do is image recognition, unmanned driving, that is computer vision (CV): how to recognize images and detect targets, these are your basic knowledge. If you are doing a recommendation system in e-commerce or content platforms, then you have to understand how the recommendation algorithm works and how to make the system “understand you better”.
You won’t have to dig deep into every of these techniques, but you must at least know what they can do and where they are limited. This allows you to define the product more clearly, communicate with the technical team in one channel, and see where there are opportunities for innovation.
Even if you are working on a general AI platform, it is useful to understand these mainstream application directions. It can help you grasp the direction of the industry, so that the product does not walk on the road and find that the direction is off.
Here are several videos of Youtube bloggers who are very professional in the field of artificial intelligence. The videos are very long, but I recommend you watch them all
2) Proficient in prompt engineering and optimization
Prompt design, writing, and optimizationBecome a core and high-frequency job content of AI product managers. Write high-quality, accurate, and effective prompts based on different business scenarios and user needs to improve the accuracy and business adaptability of model responses. It is required to establish a hint thesaurus with classified management and controllable versions. At present, the recruitment JD of some companies clearly states that it is necessary to be able to write 20 prompt words that can be commercially implemented.
Now the recruitment market for domestic AI product managers is already obvious – whether to write prompts is no longer a plus, but alsoAre you eligible to serve the table？The key to
In the past, prompt engineering ability was just a “icing on the cake” skill, but now it has become a “threshold”. Especially after AIGC is increasingly implemented in practical applications, Prompt design capabilities directly determine whether the AI products you make are smart or not and whether they are easy to use.
You can find out by looking at the recruitment requirements of major platforms
ByteDance recruited AI technology operation experts, requiring you to independently handle prompt strategies and optimize the quality and efficiency of generated content. There are also companies like Alibaba, Tencent, Baidu, Meituan, and iFLYTEK, all of which emphasize the investigation of large model understanding and prompt application capabilities in recruitment.
Prompting is not only a “pass” to help you enter a large factory, but also the key to determining whether you can get a high salary. Speaking of which, some people may ask: Prompt ability is so important, does it mean that all AI product managers have to learn the same deeply? In fact, different types of PM use different ways:
- Platform product manager: The focus is on how to encapsulate Prompt into a universal interface so that others can easily access and use it. You are equivalent to the person who “built the platform”.
- Model Product Manager: You have to work closely with the algorithm team to improve the model’s performance by tuning the prompt. You are the one who “feeds the model what to eat”.
- Application Product Manager: You are facing real users, and if the prompt is not written well, the user experience will be very poor. For example, if you make a customer service robot, the prompt can be written clearly, and it can quickly help users solve problems; If it is written vaguely, it will only “pretend to be stupid”.
But no matter what kind of AI product manager you are, you can write prompts, tune prompts, and understand the logic behind prompts.It has become a basic skill。
Here are a few recommended prompt libraries
3) Experience in the implementation of AI products or projects
The first point is that it is useless to just say that it is useless not to practice, and you have to do something with real knives and guns.
Today’s companies prefer those who have “done something with their own hands”. For example, someone in Meituan built a gadget by himself, and the result stood out. The meaning is very simple: don’t just talk about concepts and draw cakes, but a small demo that really has a project in hand and can run is the most convincing. AI products are often still in the exploration stage, and how to quickly trial and error and change quickly is the key. At this time, if you can build a running AI application prototype, you can not only save engineers’ time, but also win their trust. This style of “adjusting while doing” has actually become the default culture of AI product teams.
Second, the two most popular keywords now are RAG and Agent.
In the recruitment information, the screen is now full of words such as LangChain, RAG, Dify, Coze, and AutoGPT. Why? Because this is the core structure of the new generation of AI products. RAG can solve the problem of “not remembering things” of large models and bring in external knowledge; Agents can make the model “move” to call external tools and interact with the system, no longer just a talking robot. This kind of architecture upgrades AI products from “chat” to “real work”, which means that if you are familiar with these technical models and have practical experience, your competitiveness will be greatly improved.
The third point is that no matter how cool the technology is, it will eventually fall on “making money”.
Nowadays, enterprises require AI product managers not only to “be able to do”, but also to “be able to sell”. Companies like Weishijiajie, Liangjianghu, and Beta Technology are emphasizing whether AI products can truly become a profitable business. Don’t forget, when the AI boom first started, it was no problem to burn money; But now everyone is looking at the return on investment. Companies have spent so much money, and they must want to see real money in return. Product managers are stuck between technology, users, and business, so you must not only understand technology, but also understand how to turn a product into a marketable and profitable project. ADP said that the most important trend in 2025 is “business growth”, and Sanofi directly stated in JD that it hopes that AI product managers can “drive measurable business results”.
4) Use of AI tools
Nowadays, many companies will pay special attention to one thing when recruiting AI product managers – whether you are a “heavy user” and whether you usually like to tinker with various AI tools. In fact, there is a deeper meaning behind this, not only skilled in operation, but also reflects whether you are passionate about AI, whether you are curious, and whether you are willing to think about it yourself. This attitude of “willing to toss” is precisely what the AI industry particularly values. Because technology changes too fast, no one can rely on training and wait for others to teach, they must rely on their own learning and trying. The sooner you get hands-on with it, the better you’ll understand what these technologies can and can’t do. For enterprises, this is a sign of whether you are at the forefront of the industry or not, and it is also a rare “geek spirit”.
Not only that, AI product managers are often the first people within the company to come into contact with and promote new tools – to put it bluntly, they are the “missionaries” and “seed users” of AI. You use it first, try the effect, and then take the team to get started and use AI to optimize the process and improve efficiency. This ability to drive is no longer just about making products, but changing the way teams and even the entire company work. Like many companies that have begun to promote AI programming assistants internally, product managers are behind the promotion.
Finally, AI product managers in many companies are actually becoming more and more like “makers” – not just to mention requirements, but to build prototypes by themselves, and even directly use tools like Coze to train agents and build processes. Nowadays, low-code and zero-code tools are becoming more and more mature, and even non-programmers can handle a bunch of complex things. Whoever can get started first will be able to run faster. In the MVP stage, this capability directly determines the speed and cost of trial and error. Therefore, it doesn’t matter if you can’t write code, but you have to be able to play the tool clearly, which has become the basic quality of AI product managers.
In fact, the four core competencies that AI product managers need to master – basic understanding of AI, how to write prompts, experience in project implementation, and skillful use of various tools – are not used separately, but packaged together to work together.
For example, if you don’t understand how large models handle tokens and how long their context will be “fragmented”, then it is difficult for you to write prompts that are both cost-effective and efficient. Therefore, the theoretical basis of AI is the prerequisite for good prompt writing.
If you want to truly implement an AI project, it is not enough to know how to write prompts. For example, if you want to create an intelligent customer service, you must first design a reasonable conversation flow (this depends on prompt engineering), and then use a platform like Coze to repeatedly try and optimize (this depends on tool proficiency).
Moreover, people who play with tools have another advantage – they can detect new trends in the industry faster. For example, if you have just started a new multimodal model and find that it has a bright spot, then you may be inspired to come up with a new application scenario. Then judge whether it is technically reliable, which uses the basic knowledge of AI.
Therefore, instead of examining whether you can write prompts and use tools, companies pay more attention to whether you can connect these things in series to solve practical problems. For example, an excellent AI product manager knows which scenarios large models are prone to “nonsense”, so through exquisitely written prompts, coupled with architectures like RAG, and using LangChain to build a Q&A system, it can really help users find the exact information they want – this is hard power.
