A Beginner's and Dissuasion Guide to Becoming an AI Product Manager
-
This article summarizes the essential knowledge required before entering the field of AI product management and the preliminary cognitive framework to establish.
"Artificial intelligence is the future, and the future is already here." This rhetoric has become widespread and deeply ingrained.
Many product managers or newcomers who haven't yet entered the industry are debating whether to become AI product managers, as the term "AI" carries a glamorous and cutting-edge connotation, making it seem like they can ride the wave of the era's benefits.
So before diving into the knowledge about AI product managers, let's first dissuade some.
A couple of years ago, the AI industry was booming, but it has since cooled down significantly. However, the market is still flooded with various AI training courses or Python classes, as if not learning Python means falling behind the times.
This raises two issues: first, whether learning Python is genuinely useful for oneself, and second, whether the AI industry truly has the limitless prospects people claim.
Let's address the first issue:
These training courses are essentially creating a sense of anxiety and capitalizing on it to sell their programs.
It's not that learning Python or acquiring a new skill is useless, but rather that one should align their learning goals with their actual needs and circumstances.
Otherwise, blindly following trends may result in wasted time, money, and a painful learning process, only to end up with superficial knowledge or skills that don't contribute to one's work.
As for the industry's development, if someone is determined to dive into the AI industry to catch the so-called "era benefits" and seek job opportunities, such as becoming an AI product manager, the advice remains: don't.
Information dissemination has a delay. The current state and cutting-edge developments of a niche industry are first known by its practitioners, such as those working in AI-focused companies, followed by investors who closely monitor industry trends, and finally the media.
By the time information reaches the media after multiple rounds of dissemination, it is often severely distorted and outdated.
So, when you see overwhelming media coverage praising an industry, the reality has likely changed drastically—either the favorable market conditions have vanished, or the market landscape has already solidified.
If you're not already in the industry and only learn about a "hot trend" through media or hearsay, don't expect to ride the wave of the era's benefits. Jumping in now would only make you a target for exploitation, much like those rushing to enroll in Python courses.
So, how can one truly gauge an industry's prospects?
There are two methods:
The same applies to the AI industry. To understand its current state, either join a leading company or talk directly to practitioners.
Recently, I came across an amusing "anecdote" on Zhihu:
Someone asked: Which industries will AI replace in the future, and will it cause mass unemployment?
Many responded with lengthy analyses about the difficulty of replacing various industries, but one answer stood out: "I don't know which industries AI will replace or how much unemployment it will cause, but right now, many AI practitioners are losing their jobs."
This reflects the true state of the AI industry. As the hype fades and capital withdraws, it's clear who was swimming naked.
The biggest challenge facing the AI industry is: difficulty in implementation and finding suitable application scenarios.
Burning money to capture the market is unsustainable. The core issue is whether real user needs can be identified and genuinely addressed.
A company's ability to generate positive cash flow determines its survival during a capital winter.
A profitable business isn't necessarily a good one, but at least it's not a bad one.
Profitability means self-sufficiency, indicating it's not a bad business. However, whether it's a good business depends on market size (i.e., the ceiling) and scalability, where marginal costs decrease during expansion.
Most projects pitched with flashy PowerPoints are based on pseudo-demand. Currently, AI has found successful applications in industries with proven demand, such as security and finance.
The conclusion is: It's not advisable for product managers or newcomers from other industries to enter the AI field now.
It's better to start in mature industries like e-commerce, finance, or content communities.
Take autonomous driving as an example. It may indeed be the future, but widespread adoption is unlikely in the next 5 to 10 years. Becoming an autonomous driving product manager now for the sake of a first-mover advantage offers little practical learning.
Since autonomous driving is still in the experimental and testing phases, the product isn't mature, and the core lies with the technical team, leaving little room for product managers to add value.
Thus, jumping into autonomous driving product management now might leave you with little to do—there aren't enough users, scenarios, or demands yet.
However, joining a company like Didi, which has become the "utilities" of the transportation industry, is a better option. Didi's mature platform and vast travel data make it well-positioned for autonomous driving.
After gaining experience in the transportation industry, transitioning to autonomous driving later would allow you to leverage existing knowledge, giving you a greater advantage and faster progress.
For instance, when Tesla ventured into electric vehicles, Elon Musk recruited heavily from the traditional automotive industry, innovating on an established foundation.
Technological progress is always continuous, building on existing advancements until a tipping point transforms quantity into quality.
If, after this dissuasion, your goal remains unchanged and you're fully prepared, then entering the AI industry as a product manager is still a great choice.
Because no matter what you do, conviction is the first step—believing in yourself is key to success.
Being a product manager in the AI industry isn't fundamentally different from other industries; the methodologies are similar. The main distinction is the stronger emphasis on practical implementation in AI.
Below are 10 questions I compiled while reading The AI Product Manager, which can help establish a preliminary cognitive framework for aspiring AI product managers.
The internet primarily restructures production factors (i.e., business models), while AI upgrades them.
For example, in transportation, platforms directly connect drivers and passengers, restructuring online and offline processes. AI, however, redefines the vehicles and drivers themselves through autonomous driving technology.
At its core, AI products revolve around the concept of probability.
Using historical data to calculate the likelihood of future events is essentially predicting the future.
Every inference and prediction made by AI comes with uncertainty about the outcome.
The reason AI hasn't been widely adopted is that expectations for its probabilistic performance vary across industries and scenarios, leading to uneven adoption rates.
For instance, in healthcare, an 85% diagnostic accuracy rate for AI isn't enough for widespread use, but 99.99% would be.
However, one must consider the sample size when calculating probabilities.
After several Tesla autopilot accidents, Elon Musk tweeted that autonomous driving is far safer than human driving, with accident rates significantly lower than the overall societal average.
This seems to align with people's intuition—Tesla's traffic accidents might only be in single digits, but road accidents happen every day. Are we being too harsh on new technology?
In fact, Musk's statement makes a mistake: when discussing Tesla's accident rate, the sample size is very small, whereas the overall traffic accident rate is based on the vast number of vehicles and miles driven across society. Tesla's accident rate could double with just a few more incidents.
Therefore, the current sample size makes Tesla's accident rate calculation inaccurate. We cannot directly conclude that autonomous driving has a lower accident rate than human driving.
Only when the number of Tesla's autonomous vehicles and miles driven accumulates to a certain scale, with a sufficiently large sample, can we compare it with human driving accident rates and truly determine whether autonomous driving is superior.
Product managers designing AI products must balance probability optimization and cost investment.
They need to assess the probability of meeting user needs, the minimum acceptable standards for users, and the standards that exceed user expectations. Based on these assessments, they should decide on the product development investment strategy.
Don’t pursue perfection—commercial success is the top priority.
The three core elements of AI are: algorithms, computing power, and data.
AI products should also consider these three elements:
(1) Algorithm Level: Product managers should have a basic understanding of mainstream algorithm models and frameworks, and be able to quantitatively evaluate their effectiveness in different scenarios.
(2) Computing Power Level: Product managers should start from demand, assess the system architecture required to support the algorithm models for product features, evaluate hardware costs, and decide whether to use Platform-as-a-Service or build their own computing platform.
(3) Data Level: From the outset, product managers must consider where the data comes from, how to ensure data quality, and how to implement data governance.
The competitive edge of AI products lies at the intersection of these three elements.
Core technology, productization, and commercialization are indispensable for the success of an AI product.
(1) Core Technology: Core technology is the primary factor for an AI product's success.
(2) Productization: Promote product value (reach users in a fast, low-barrier way), quickly demonstrate value (strategies to help users understand and be convinced by the product), deliver value (ensure stable, long-term value delivery), and extend value (make users rely on the product and integrate it into their lives).
(3) Commercialization: Productization determines the product's value potential, while commercialization determines its ability to monetize that value.
AI commercialization requires product managers to thoroughly analyze scenarios and pain points, assess the product's value and R&D costs, and formulate suitable commercial promotion and pricing strategies—even adjusting product positioning when necessary—to achieve monetization.
Balancing technical and market foresight is essential for AI product managers.
Understanding basic machine learning algorithms—their applications, problem-solving capabilities, and pros and cons—is sufficient.
(1) Product managers should deeply understand the principles and best practices of every technical action in their field and be able to explain them fluently.
(2) They should actively participate in the R&D process, such as providing high-quality training datasets for the team.
(3) Mastering cutting-edge technology applications and having insights into technological trends are crucial for designing competitive and forward-looking products.
Core AI Technologies: Natural language processing, human-computer interaction, computer vision, biometric recognition, speech recognition, virtual reality, augmented reality, mixed reality, etc.
AI Applications and Services Fall into Three Categories: Speech and text processing, image and vision, big data analysis and prediction.
(1) Industry + AI Companies: Leverage years of domain expertise to provide AI-enhanced products or services (e.g., Ford and GM developing autonomous driving tech).
For such companies, deep industry understanding and trend insights are core competencies.
(2) Applied AI Companies: Offer basic functionalities like facial or speech recognition via APIs.
Their business model is primarily B2B, with product managers focused on project payments. Thus, they need business skills (pre-sales, sales) and must differentiate between standardized and customized products.
(3) Core Tech/Platform Companies: Focus on hardware R&D, data governance, and AI modeling for underlying platforms.
Here, product managers emphasize understanding foundational technical frameworks.