Your Interviewer Just Asked: 'AI Writes 80% of the Code Now — Why Do We Still Need You?'

This question is becoming more and more frequent

Let's start with a fact: technical interviews in 2026 are completely different from two years ago.

Two years ago, interview questions were: "Write a Promise by hand," "Explain the principles of React Fiber," "What is the browser rendering process?"

Now, interviewers assume you can use AI. What they really want to know is: In an era where AI can write code, what is your irreplaceability?

I asked around in a tech community, and this year there are at least three ways they ask:

• "AI can write 80% of the code, so what's your value?"
• "If I gave you an intern plus Claude Code, could they replace you?"
• "What's the division of labor between you and AI?"

Essentially, it's the same question. Answer it poorly, and you're out.

Most people's answers fall into traps

I collected some common answers, and interviewers are generally unsatisfied:

❌ Answer 1: "The code quality from AI isn't good enough; humans still need to write it"

This answer worked in 2024. It doesn't work in 2026.

Claude Code and Codex already generate code of quite high quality. Most CRUD interfaces, form components, and utility functions are written better by AI than many junior developers. If you say "AI code quality is poor," the interviewer thinks: Doesn't that mean your skill level is about the same as AI?

❌ Answer 2: "AI doesn't understand business requirements"

The interviewer will follow up: "If the product manager writes the requirements clearly, can't AI understand them?"

You'll find it hard to argue. Because the fact is—most requirements can indeed be described clearly in natural language, and AI can indeed generate code based on those descriptions.

❌ Answer 3: "Someone still needs to do Code Review"

This answer positions yourself as an "AI quality inspector." The interviewer will think: A quality inspector's salary doesn't need to be 20,000.

How I later answered

After being asked four times, I figured out one thing: This question isn't testing your attitude toward AI; it's testing your understanding of your own value.

My current answer has three layers:

Layer 1: AI writes code; humans make decisions

"AI can write 80% of the code, but it can't write that 20% of decisions."

Here's a concrete example. Last month, we were working on an e-commerce campaign page. The requirement was "show a countdown after the user places an order."

AI can perfectly write a countdown component. But it won't ask you these questions:

• What happens when the countdown ends and the user's page is still open? Auto-refresh or pop-up notification?
• If the user modifies their local time, will the countdown be bypassed?
• Under high concurrency, if tens of thousands of countdowns hit zero simultaneously, can the backend handle it? Should the frontend implement request queuing?
• Does this countdown need to sync with server time? What if the client time is inaccurate?

Each of these questions could lead to a production incident. AI won't proactively think of them, because AI only solves "the problems you raise," not "the problems you didn't think of."

The value of a senior developer isn't writing code; it's knowing which code shouldn't be written, which scenarios will cause trouble, and which decisions will affect maintenance costs for the next six months.

Layer 2: AI can write a file; humans can design a system

"AI is an extremely excellent executor, but it lacks a system perspective."

Ask Claude Code to write a user registration endpoint, and it will do a great job. But ask it to design the entire user system, and it doesn't know:

• Should registration and login be split into two microservices?
• How should user data be sharded? By user_id hash or by registration time range?
• Should sessions use JWT or Redis? What are the trade-offs for each?
• If we need to integrate third-party login (WeChat, Google) in the future, should the current table schema reserve extension fields?

These are architectural decisions that require comprehensive judgment based on business scale, team capability, current tech stack, and future plans.

AI can list five options for you, but it doesn't know which option suits your company. Only a human can make that judgment.

During an interview, I said something that made the interviewer nod:

"AI lowers the barrier to writing code, but it raises the barrier to making the right decisions. Because code generation is so fast now, a wrong decision will turn into a mountain of technical debt faster than ever."

Layer 3: AI cannot be held accountable for results

"When a production incident happens, AI won't be woken up by on-call."

This layer sounds like a joke, but it's the most fundamental.

Code is deployed to production. At 3 AM, an alert goes off. Someone needs to:

• Assess the impact scope
• Decide whether to roll back
• Coordinate frontend, backend, and operations teams for troubleshooting
• Deliver a fix within half an hour
• Write a postmortem report and drive process improvements

Every single one of these tasks requires judgment, communication skills, and a sense of responsibility. AI can help you check logs and analyze stack traces, but it can't make decisions or bear responsibility.

A company paying you 20,000 a month isn't buying your time to write code; it's buying your judgment and sense of responsibility.

What the interviewer really wants to hear

To summarize, the correct structure for answering this question is:

Layer	Core Idea	One-liner
Execution	AI writes code, humans make decisions	AI solves problems you raise, but won't discover problems you didn't think of
Architecture	AI writes files, humans design systems	AI can list options, but doesn't know which one fits your company
Accountability	AI can't be woken up to fix bugs	The company isn't buying code; it's buying judgment and responsibility

Finally, a bonus move: Give a real case.

Don't vaguely say "AI isn't good enough." Tell a scenario you personally experienced:

"Last month, AI helped me write a data export function, and it ran perfectly. But during my review, I noticed it didn't implement pagination—100,000 records loaded all at once. It was fine in the test environment, but in production it would have caused an OOM. AI wouldn't realize this problem because it doesn't know how large your production data volume is."

One concrete case is worth ten correct but empty statements.

The core competitiveness of frontend engineers in the AI era

If you're still agonizing over "should I learn AI or not," you're asking the wrong question. AI is a tool, not a competitor.

The real question should be: What skills become more valuable as AI gets stronger?

1. System design ability — The faster AI generates code, the higher the cost of making wrong design decisions
2. Business understanding — Understanding the "why" behind requirements, not just the "what"
3. Cross-team collaboration — Coordinating frontend, backend, product, and design; this can't be solved by writing code
4. Production incident response — Being able to handle incidents, investigate, fix them, and prevent them from recurring

These four skills become scarcer as AI gets stronger.

Finally

Next time you're in an interview and get asked "What use are you now that AI can write code?", don't panic.

This question isn't questioning you; it's giving you an opportunity to demonstrate high-level thinking. The people who can answer this question well are precisely the most valuable people in the AI era.

Have you been asked a similar question in an interview? How did you answer? Share in the comments.