A Frontend Friend Was Laid Off, the Boss Said 'One Person Using AI Can Replace Three.' I Decided to Test Which Jobs Are Actually at Risk.
2 months ago
A friend of mine was laid off last month. He's a frontend developer, not bad at his job. He went to ask the boss: 'Is it me, or is it the company?' The boss said: 'Last year, one person using AI could replace three. It's not you, it's your position.'
This made me decide to figure out one thing: which positions are actually experiencing this?
Everyone is talking about how many jobs AI will replace: Citi says 70 million jobs are at risk, McKinsey says 30% of jobs will be replaced. Big numbers, but do you feel they relate to you?
I personally tested 20 positions, broke down the core tasks of each, and checked item by item whether AI can do them, how well, and where it falls short. I designed a 4-dimension scoring framework and calculated a replacement index for each position.
This index doesn't mean a certain job will definitely disappear, but it can tell you the direction of the wind. Hope it helps.
4 Dimensions to Clarify Whether a Job Is at Risk
Asking 'Can AI do this job?' is too broad. It needs to be broken down.
Dimension 1: Task Standardization (1-10 points)
How much of what you do daily follows a fixed process with standard answers?
Data entry, invoice review, customer service order checking—every step has an SOP. AI does these faster and with fewer errors. Score high.
Psychological counseling, brand positioning—no standard answers, every judgment is different. Score low.
Dimension 2: Judgment Replaceability (1-10 points)
How much of your work involves making decisions? Do those decisions have clear right or wrong answers?
The higher the score, the more rule-based the judgment, the easier for AI to replace.
Customer service answering questions with scripts—clear rules, little judgment. Score high.
Product manager deciding feature priorities—no standard answer, must be accountable for consequences. Score low.
Dimension 3: Interpersonal Interaction Density (1-10 points)
How much does your work effectiveness depend on real trust and empathy between people?
Backend programmer, data annotator—work with computers, almost no human interaction. Score high.
Sales director, psychologist—clients buy you, not the data you output. Score low.
Dimension 4: Accountability Clarity (1-10 points)
If something goes wrong, who can be held responsible?
Translation has original text for reference, right or wrong can be objectively judged, humans can fix AI errors. Score high.
Surgeon, trial lawyer—if AI makes a mistake, who is responsible? Would you tell the patient 'It was AI's fault'? Score low.
AI Replacement Index for 20 Positions: Full Ranking
| Rank | Position | Standardization | Judgment Replaceability | Interpersonal Density | Accountability | Composite Index |
|---|---|---|---|---|---|---|
| 1 | Data Entry Clerk | 10 | 9 | 9 | 9 | 93 |
| 2 | Telemarketer/Cold Caller | 9 | 8 | 8 | 8 | 83 |
| 3 | Basic Translator | 9 | 6 | 9 | 8 | 80 |
| 4 | Basic Accountant/Bookkeeper | 8 | 7 | 8 | 7 | 75 |
| 5 | Basic Customer Service | 8 | 8 | 6 | 7 | 73 |
| 6 | Warehouse Manager | 8 | 6 | 8 | 6 | 70 |
| 7 | HR Resume Screening | 7 | 6 | 5 | 6 | 65 |
| 8 | Junior Programmer | 7 | 5 | 7 | 6 | 63 |
| 9 | Quality Inspector | 7 | 6 | 7 | 5 | 63 |
| 10 | New Media Editor (News) | 7 | 5 | 4 | 6 | 55 |
| 11 | Graphic Designer (Template-based) | 6 | 4 | 4 | 6 | 50 |
| 12 | Bank Loan Approver | 6 | 3 | 5 | 4 | 45 |
| 13 | Data Analyst | 5 | 3 | 5 | 5 | 45 |
| 14 | Translator (Legal/Literary) | 4 | 2 | 2 | 3 | 28 |
| 15 | Sales (B2B Key Account) | 3 | 3 | 1 | 3 | 25 |
| 16 | Product Manager | 3 | 1 | 2 | 3 | 23 |
| 17 | Electrician/Field Maintenance | 1 | 2 | 4 | 1 | 20 |
| 18 | AI Solutions Architect | 2 | 1 | 2 | 2 | 18 |
| 19 | Psychologist | 2 | 1 | 1 | 2 | 15 |
| 20 | Surgeon | 1 | 1 | 1 | 3 | 13 |
Composite Index = Average of 4 dimensions x 10. Higher means more at risk.
Next, Analyze High-Risk Positions One by One
No.1 Data Entry Clerk · Replacement Index 93
This job isn't 'on the way.' It's already at the end.
Replacement rate has exceeded 95%.
A finance person at a logistics company shared: Previously, month-end reconciliation took 3 people 3 days. Now, an AI financial tool does it in 30 minutes with higher accuracy.
An accounting firm: 3 data entry positions cut to 2, the remaining one mainly handles exceptions and tax planning, no longer related to 'entry.'
AI is perfect for this: invoice recognition, expense report review, report generation—all standardized 'input-process-output' workflows. AI is not only faster but also more accurate than humans.
What can humans still do? Almost nothing.
The essence of this job is 'using humans as machines.' When real machines arrive, humans are no longer needed.
No.2 Telemarketer/Cold Caller · Replacement Index 83
Replacement rate exceeds 90%.
AI voice bots have basically taken over most outbound marketing. The sales calls you receive now are likely not from real people.
Outbound scripts are highly standardized. Opening lines, product introductions, objection handling—all in the SOP. AI voice bots sound indistinguishable from real people, and they don't get tired, hang up, or get yelled at.
High-net-worth client negotiations with strong emotions still need real people. When someone says 'Get me your manager,' that's when a human steps in.
No.3 Basic Translator (Technical Docs/Business Letters) · Replacement Index 80
The latest generation of large models has pushed general translation quality to near-human levels.
A translation company owner revealed: Simple translation business decreased by 60%.
Microsoft's 2025 research shows: Interpreters and translators rank #1 in AI exposure.
Technical documents, business emails, product manuals—these texts have fixed terminology and sentence structures, exactly the pattern matching AI excels at. Cost is 1/10 of human translation, speed is in seconds.
Literary translation requires capturing feeling. Brand localization requires cultural nuance. Legal contract translation: a single word deviation could be worth millions.
Demand for high-end translation is actually rising, but the threshold for 'high-end' has changed.
No.4 Basic Accountant/Cashier/Bookkeeper · Replacement Index 75
At one accounting firm, AI has completed 97% of voucher and tax filing work. Recruitment demand for junior bookkeeping positions has halved year-on-year.
AI financial robots can now: invoice recognition, bookkeeping, tax filing, reconciliation, payroll calculation, report generation. Full-process closed loop, no human needed.
The underlying logic of accounting is processing data according to rules. Accounting standards are rule sets, source documents are data sources. Having a human calculate 1000 invoices according to tax law takes a day; AI takes 30 seconds.
But tax planning, audit communication, financial strategy—these require judgment, not execution.
Basic accounting is disappearing, demand for financial analysts is rising.
No.5 Basic Customer Service · Replacement Index 73
Intelligent customer service can already handle over 80% of standard inquiries—order tracking, address changes, return/exchange processes—all in SOPs.
The remaining complex complaints and emotionally charged situations require human intervention.
A bank piloting an unmanned customer service center: reduced from 200 to 8 people.
A cross-border e-commerce company in Dongguan: 50-person customer service team streamlined to 15 with AI.
The basic logic of customer service is answering questions according to procedures. AI is not only fast but also available 24/7, doesn't get tired, doesn't take leave, doesn't break down from being yelled at. For most standard scenarios, customers can't tell if they're talking to a human or AI.
But high-difficulty complaints with strong emotions—AI can't handle them. When someone says 'Get me your manager,' what's needed is a real person with decision-making power and empathy. AI actually knows how to handle it, but customers don't believe AI has that authority.
Special Attention: Junior Programmer · Replacement Index 63 (Rank 8)
Although not in the top 5, the speed and scope of change in this position deserve separate mention.
Over 25% of new code at Google is generated by AI.
GitHub Copilot increases individual engineer efficiency by 30%-50%.
In plain English: 3 programmers who use AI do the work of 5 people before.
AI hasn't replaced programmers, but it has raised the bar. Pure frontend job opportunities are decreasing, and positions for newcomers with less than 1 year of experience have significantly shrunk.
Microsoft lists 'Web Developer' among the top 40 occupations with highest AI exposure.
CRUD, unit tests, boilerplate code—these have clear patterns and best practices, and AI-generated quality is already at a deployable level.
AI coding tools like Cursor allow what used to take junior programmers most of a day to be done in ten-plus minutes.
What does AI still lack? Architecture design, business logic decomposition, complex system debugging.
Jensen Huang once said: 'You won't be replaced by AI, but you will be replaced by someone who uses AI.' For programmers, this is no longer a prediction; it's happening.
These 5 Positions Are Actually Becoming More Valuable
Looking at the rankings above might feel discouraging. But at the same time, some positions' demand and salaries are rising, not falling.
Electricians and Blue-Collar Technical Workers
Surprisingly leading, electrician was the occupation with the highest employment growth last year.
The massive global construction of data centers (AI's underlying infrastructure) has driven up demand for electrical installation, maintenance, and inspection. These require on-site hands-on operation, and embodied intelligent robots haven't reached that level yet.
B2B Key Account Sales
In a survey by Cheung Kong Graduate School of Business, sales ranked #1 as the 'safest job in the next 3-5 years.'
Customer trust is built bit by bit; AI-generated scripts can't replace that. A B2B big deal involves relationship maintenance, needs discovery, multi-party negotiation—these require humans to take responsibility and make decisions.
AI Solutions Architect
Annual salary up to 1.5 million RMB. According to the 2026 salary report by Career International, embodied intelligence algorithm engineers can earn up to 2 million RMB annually, and AI solutions architects up to 1.5 million.
They're not avoiding AI; they're harnessing it. The market is telling you with real money: in the AI era, knowing how to use AI and understanding the industry is the real currency.
Psychologist
Replacement index only 15.
AI can actually do it—Claude and GPT can simulate empathy and do it convincingly. But what clients need is a real person listening. Trust is rooted in 'the other person is also human.' AI can't cross that line.
Surgeon
Replacement index only 13, the lowest of all.
Technically, it's already possible—the Da Vinci surgical robot can perform fine operations without issue. The real problem is: if something goes wrong during surgery, who is responsible? Try telling the family 'It was AI's fault.'
The Pattern: AI Kills Standardizable Tasks, Not Entire Jobs
After evaluating these 20 positions, I found something.
AI doesn't kill jobs; it kills the parts of jobs that can be standardized and templated.
| In your daily work | Which zone are you in | What happens |
|---|---|---|
| Over 80% follows clear rules | High-risk zone | Being replaced by AI in batches |
| 50% rules + 50% judgment | Transition zone | Job still exists, but requirements have changed |
| Over 80% requires decision-making, empathy, or hands-on work | Safe zone | Becoming more valuable |
The accounting profession hasn't disappeared, but basic bookkeeping is rapidly shrinking, while demand for tax planning is rising.
Programmers haven't disappeared either; what's disappearing are those who only write code. Those who can define requirements and orchestrate development with AI are being fought over.
Translation is the same: the profession still exists, but technical document translation has been largely eaten by AI, while high-end brand localization translators have seen their value increase.
Under the same job title, those who execute and those who judge are heading toward completely different fates.
Test Your Own Position
Use my 4-dimension framework to score your position. Each item 1-10 points, higher score means more at risk:
| Dimension | Ask yourself this question | Your score |
|---|---|---|
| Task Standardization | 'How much of what I do daily has a fixed process and standard answers?' | ___ / 10 |
| Judgment Replaceability | 'How many of my decisions have no right or wrong answer and require a call?' (reverse) | ___ / 10 |
| Interpersonal Density | 'How much does my work effectiveness depend on real interpersonal trust and empathy?' (reverse) | ___ / 10 |
| Accountability | 'If something goes wrong in my work, can AI take responsibility for it?' | ___ / 10 |
Add total, divide by 4, multiply by 10 = Your Replacement Index.
- 0-30: Your job has deep barriers; AI won't touch you in the short term.
- 30-60: You're in the transition zone. The job title may stay the same, but skill requirements have changed. You need to shift from executor to AI harnesser.
- 60+: What you do, AI can already do or is learning. I'm not fear-mongering; this is a signal you need to face.
When I wrote this post, I also plugged in a few positions I've held before. The results were mixed...
In the same position, those who execute are replaced, those who judge become more valuable.
My back felt a bit cold after reading this. I'm that frontend developer who's been working for three years. Last year I was praised for writing good code, but this year a newcomer using Cursor joined the team, doing twice my previous daily workload. The boss didn't fire me, but asked me to transfer to AI code review, essentially becoming a quality inspector for AI. The article mentions a "63% replacement index for junior programmers" – I think that score is too low. At least in my team, positions for new graduates have been frozen for almost a year. But on the flip side, I've become much faster at writing code with AI now. Maybe it's true what Jensen Huang said: you won't be replaced by AI, but by someone who knows how to use AI.