Short answer: Give the same AI and the same project to two people, and a veteran PM can pull increasingly valuable things out of the model, while a newcomer holding the strongest model still spins in place. The gap is not the tool. It is whether you have a complete "project sandbox" in your head. AI will not make you stronger by itself; it only amplifies the judgment you already have. There are three practices: translate an execution action into a logic slice, build the judgment that input determines output, and deliberately train adversarial thinking.
The gap is not the tool
Have you noticed this: with the same AI and the same project, a PM who has been doing the work for years seems to get more and more valuable output from the model, while many new PMs, even with the strongest model in their hands, get results that fall short.
The gap really is not the tool. It is whether you have a complete "project sandbox" in your head.
AI will not directly make you stronger. It will only amplify what you already have. Think of AI as the same dragon-slaying sword handed to everyone on the project. Give it to a newcomer who has not yet mapped the whole project, and a few swings later they are still chopping at the work right in front of them. Give the same sword to a veteran who already holds the rhythm of the whole version in their head, and one swing can straighten out an entire line.
AI itself has no experience. It is more like a giant puzzle that knows a little about everything, but what it is assembled into depends on the sandbox in your head. The depth of the questions you can ask is the depth of the answers it can give you. The more precise your judgment of version rhythm and team collaboration, the more valuable it is when AI helps you find gaps.
Conversely, if someone has no sandbox in their head, even the strongest model will only help them spin in place faster.
How people in execution roles can build the sandbox
Many people will say: I am in an execution role right now. I write requirements and chase progress every day. Where am I supposed to get the chance to build a sandbox?
The chance is around you every day. Starting today, treat AI as a review partner that is always on call, and deliberately practice three things.
First, translate an execution action into a "logic slice."
Most people are staring at "when will the work in my hands be done." What you should be asking is "where does this work sit in the whole project, and what is it blocking."
The method is simple. After you finish a schedule, do not rush to send it out. Feed it to AI first — tell it you are the PM for this area, you just wrote this schedule, and ask it to stand, in turn, in the shoes of the producer, the lead engineer, publishing, and even a core player, and say what this schedule means to each of them, what the risks are, and what it means for the rhythm of the whole project.
Practice this long enough, and you will find you are no longer just a screw staring at your own module. You become the person who holds the whole project in their head.
Second, build the judgment that "input determines output."
Many people treat the process documents they wrote as treasures, and even want to quietly copy them when they leave a job. But in the AI era, the artifacts themselves are no longer what is valuable. What was always valuable is the chain of logic behind the document: the few boundaries and ground rules you decided on.
Third, deliberately practice "adversarial thinking."
When you want a decision recommendation, do not let AI just agree with you. Have it switch identities and push back, and force it to pick out the weakest link in your plan. The more you get pushed, the sturdier your own judgment becomes.
The watershed
These three practices are not AI tricks. They are the habit of examining your own work.
Practice them long enough, and you gradually change from "a person who executes" into "a person who examines their own work." That is precisely the watershed that pulls you apart from other executors. Everyone can have the tools. Not everyone has the sandbox.
That is also why the same model produces such different results in different hands. The tool does not close the gap. The sandbox in your head does — and the sandbox is built by how you look at your own work, not by which model you open.
This English version is adapted from the public Chinese original: https://pmnote.ai/articles/bilibili-bv1cajv6teds/
More notes on game project management at https://pmnote.ai, and in the WeChat mini program “Hao的游戏PM笔记”.