"The proposal is ready. Everyone, please take a look. We'll go through it tomorrow morning."
Immediately after, a document of over twenty pages was shared.
With a complete title, a clear table of contents, and sections on industry trends, user analysis, competitive benchmarking, and an execution plan, at first glance, it looked like a rather "professional" proposal.
But as the person responsible for its implementation, you spotted problems in less than ten minutes of reading.
The market data cited in the proposal had no traceable sources. A competitor was credited with a feature that didn't actually exist. The so-called "user pain points" were just a few generic platitudes applicable to any industry. Most importantly, there was no mention of budget, timeline, or staffing arrangements.
The next day in the meeting, the author of the proposal said, "This is just a first draft. The specific details still need everyone's input."
At that moment, you realized the other person hadn't completed a proposal; they had used AI to create a new team task.
01
In the past, we often judged whether a task was complete by checking if there was a deliverable. Now, that standard has changed.
With AI, anyone can generate a structurally complete report, a professionally worded proposal, or a seemingly information-rich PPT in ten minutes.
The problem is, "looking complete" is not the same as "actually being complete."
There's a term abroad called "Workslop," which can be understood as "workplace AI garbage"—superficially complete work output rapidly generated using AI but lacking fact-checking, professional judgment, and actual value.
The most troublesome part isn't just the poor quality, but that it transfers the work cost to others.
The generator saves two hours, but the receiver spends four hours verifying information, understanding intent, and correcting errors.
One person's efficiency improves, while the entire team's efficiency declines.
02
A company started encouraging employees to use AI, hoping everyone would reduce repetitive work. Xiao Lin from operations quickly found a "method to improve efficiency."
Every Friday, he would copy backend data to an AI tool to automatically generate a weekly report. The reports were quite well-written:
"User activity increased steadily this week, content strategy achieved phased results. It is recommended to continue focusing on core user needs and strengthen refined operations."
For several weeks, the manager saw no issues.
Until one time, when the company planned to increase the advertising budget based on the weekly report, the manager casually checked the raw data and discovered that the so-called "increase in activity" was simply due to a change in the statistical method.
The AI didn't know about the methodology adjustment, nor did it know about a temporary campaign that week. It merely generated a plausible-sounding explanation based on the numbers in the table.
The manager had to re-check the reports from the past month, inquiring about data sources item by item.
Xiao Lin indeed spent only ten minutes writing the report, but to determine whether this report could be trusted, the manager spent an entire afternoon.
AI saves the writer's time but consumes the reader's trust.
During the mid-year performance review, product manager Amin received a very "formal" piece of feedback:
"It is recommended to further strengthen holistic thinking, improve cross-departmental collaboration efficiency, and enhance goal orientation and result focus in complex projects."
She recognized every word, but put together, she had no idea how to improve.
Which specific collaboration had issues? What does "lack of holistic thinking" mean? To what extent must she perform in the second half of the year to count as "enhanced result focus"?
Amin later found out that the manager had simply input a few keywords into an AI tool and asked it to polish them into a piece of performance feedback.
The manager saved time organizing the language, but Amin had to spend a week guessing what the manager actually meant.
Truly effective feedback may not be eloquently phrased, but it must contain facts:
What happened? What was the impact? What specific change is expected from the person?
If these questions aren't answered, even the most professional expression merely shifts the responsibility for thinking onto the receiver.
There was another team that used AI in almost every step.
Planners used AI to generate event proposals; project managers used AI to summarize proposals; during meetings, assistants used AI to compile meeting minutes; before presentations, the lead would have AI expand the minutes into a PPT.
The entire process was very smooth, and documents kept piling up—until a client asked in a meeting, "Why do you think young users would like this feature?"
The meeting room suddenly fell silent.
The proposal stated it "aligns with the trend of young users pursuing individual expression." The PPT stated it "accurately addresses the needs of the new generation." But no one had interviewed users, nor had anyone looked at relevant data.
Everyone had touched the project, but no one had genuinely studied the problem.
In the end, the team possessed dozens of pages of documents, yet not a single person could take responsibility for the conclusions within them.
03
What AI truly amplifies is a person's work habits. Using the same AI, some people will become stronger, while others will only get better at just finishing the job.
The difference isn't in how well the prompt is written, but in whether they treat AI as an "assistant" or as someone who "takes responsibility for them."
Conscientious workers use AI to organize information, compare ideas, and check expression, then personally verify facts, supplement firsthand information, and judge whether a plan is feasible.
Those accustomed to cutting corners will submit the first draft generated by AI directly, followed by the line: "This is just a first draft; everyone can help refine it."
To judge whether a piece of work is valuable, one shouldn't just look at how fast it was completed, how beautifully formatted it is, or how many AI tools were used.
Instead, three questions should be asked:
Does it provide new facts? Does it make a clear judgment? Is there someone willing to take responsibility for the outcome?
AI can certainly help us write reports, compile meeting minutes, create PPTs, analyze data, and even propose a preliminary framework for a plan.
But it cannot understand the on-the-ground situation for us, it cannot judge pros and cons for us, and it cannot bear the consequences of our decisions.
This article is from WeChat Official Account "Liepin" (ID: liepinwang), author: Matcha Sweet and Sour Pork.
