A Chinese podcast · Same story, 3 levels
一个人工智能客服顶了七百个人，平均满意度和人一样：一年后公司把人招了回来，因为平均值把最糟的那批客户藏起来了
One AI Did the Work of Seven Hundred Agents and Scored the Same on Average — a Year Later the Company Hired the Humans Back, Because the Average Was Hiding the Worst Customers
About this story
Klarna launched an AI customer service agent in February 2024: 2.3 million conversations in month one, the work of about 700 agents, two minutes per query against eleven. By mid-2025 it was rehiring humans. Average satisfaction had looked level the whole time, because the distribution was bimodal. Chinese listening practice at four levels. HSK 2 Chinese listening practice.
This is an HSK 2 Chinese listening episode that runs about 3 minutes. The full Mandarin script is shown with tap-for-pinyin and a line-by-line English translation, so you can listen and read at once — comprehensible input in the sense of Stephen Krashen's i+1 theory. It teaches 12 key vocabulary words such as 平均、情况、简单 and walks through 3 grammar patterns, each explained in English with examples. The same news story is retold at 3 difficulty levels — use the level selector above to find the version that is challenging but still understandable for you.
Read at your level
原文Read the complete story in Chinese. Reveal pinyin and English only when you need them.
English transcript reference
Today's story is about something that played out over two years.
The whole thing is full of numbers, and the numbers misled people.
First, the company.
It is a Swedish company, and what it does is payments.
You buy something online, and you can pay for it later.
A lot of people use it, so a lot of people have questions.
That means the company needs a lot of customer service staff.
Customer service means the people who answer your messages.
This company had seven hundred of them.
In February 2024, the company built something new.
It was a customer service agent made with artificial intelligence.
In its first month, it handled 2.3 million conversations.
The company said that was the workload of seven hundred agents.
And a few of the other numbers looked even better.
A person took an average of eleven minutes per reply.
The machine took two minutes.
And far fewer people had to ask the same question twice.
The company said it would make a lot more money that year.
When those numbers came out, other companies were watching.
Around the same time, another company put up a huge billboard in the street.
It read: stop hiring humans.
It sounded as though the matter was settled.
But a year later, things changed.
In 2025, this company started hiring again.
Hiring customer service staff. Human ones.
The company's boss said one thing.
He said: we went too far.
He said: we looked too hard at cost.
He said: the result was that quality dropped.
The interesting part here is how "quality dropped" got noticed.
The company had been watching one number: whether customers were satisfied.
On average, the machine and the humans were about the same.
On average, you could not see the problem.
But the average put two different situations together.
For simple questions, the machine replied fast and well, and customers were very satisfied.
For complicated questions, the machine also replied fast, but the reply was wrong.
Those people were extremely unsatisfied.
One end very high, one end very low, and the average came out at "about the same".
The company was looking at the number in the middle, so for a long time it saw nothing.
Now they have people and machines working together.
Simple questions go to the machine, complicated ones go to a person.
The way they hire has changed too: you work when you have time.
Among the people doing it are students, and some of the company's own customers.
What makes this worth telling is not that machines are no good.
Machines answer simple questions far faster than people. That is true.
What is worth telling is that average.
It was not a fake number. It was calculated exactly as stated.
But it put the least satisfied customers behind a good-looking figure.
Think about it:
Someone shows you an average and tells you things are fine.
What is the next question you ask?
Listen again
Try it without the transcript and notice what sounds clearer.
What vocabulary does this episode teach?
词汇HSK 5. The subject of the episode, and the thing that hid the problem for a year.
HSK 4. 平均把两种情况放在了一起 — two situations, one number.
HSK 3. 简单的交给机器 — the split they ended up with.
HSK 4. Where the machine answered just as fast, and answered wrong.
HSK 4. 结果是质量下来了 — the boss's own word for what happened.
HSK 6. 我们太看成本了 — what they were optimising, and it worked.
就是回消息的人 — defined in the episode before it is used.
The neutral word the episode uses instead of arguing about AI.
The people on the other end, split into two very different groups.
The metric on the dashboard, and the reason it stayed green.
Named once, then dropped — the episode is not really about it.
HSK 5. What went down while every efficiency number went up.
* beyond level超纲词
What grammar patterns appear in this episode?
语法又……又……
Both X and Y at once. Two qualities holding simultaneously — here, fast and good, which was true for half the cases.
简单的问题，机器回得又快又好，客户很满意。
一边……，一边……，平均下来……
One side is X, the other is Y, and averaged out… A plain-language way to describe a bimodal distribution without the word.
一边很高，一边很低，平均下来正好是"差不多"。
不是因为……
Not because of X. Used to head off the obvious reading before offering the real one.
我觉得这件事值得说，不是因为机器不行。
Proper Nouns
专有名词Sources
来源Free account
Keep learning from this story
Create a free account to keep saved words and your preferred level together.
- Keep words with their story context
- Remember your preferred level
- Build your vocabulary over time