Will AI Take My Job? A 2-Minute Test for Developers

Someone asked me the other day, “Will AI take my job?”

I asked them, “What’s the delta you’re bringing to the table right now?”

They said five years of experience. So I asked them to name the hardest decision they had made themselves in those five years. Not one the architect made, not one the client made. They couldn’t name one.

They had closed every ticket they were given, on time, for five years. Someone else had made the decisions in those tickets before they ever reached them.

So, will AI replace software engineers?

AI will replace work that is fully specified before it reaches a developer. Engineers whose week is mostly that kind of work are the ones at risk, and in Indian IT services that’s a lot of people.

Give AI a clear ticket and a pattern to copy, and it writes the code faster than you or me. Tickets, change requests, CRUD screens and migrations someone else planned make up most of what a service project hands you.

Look at how much of your work was decided before it reached you. Whatever is left after you take that away is your delta. It isn’t your fault, but it is your problem.

The test

It has two parts. Keep your ticket tracker open and give it two honest minutes.

Part 1: your last 10 tickets

Open Jira, or whatever your team uses, and pull up the last 10 tickets you closed. For each one, ask yourself:

Did I make a decision here that nobody handed me?

Count it as a yes only if you did something like this:

• picked the schema, the index or the approach yourself
• found the root cause of a bug that the ticket didn’t describe
• pushed back on a requirement and got it changed
• decided what not to build

Copying the pattern from the last module is a no, and so is following the design doc or doing exactly what the client spec said, even if it was hard work and took three days.

Score 1 point for every yes. That’s 0 to 10.

Part 2: three interview questions

Answer each one out loud, as if an interviewer just asked it. Count it only if you could explain it in a minute without guessing.

1. Which is the slowest API in your service, and why is it slow?
2. Your biggest table grows 10× overnight. What breaks first?
3. You need to rename a field in an API response. Which clients break, and how do you ship the change without breaking them?

Score 2 points for every answer you could explain. That’s 0 to 6.

Add both parts. Your total is out of 16.

What your score means

0 to 5: AI can already do most of your week

Most of your work arrived fully specified, and that’s the work AI does best. It says more about your project than about you, but the person deciding headcount won’t see the difference.

Start with one problem where you have to decide yourself. The 1 Million Row Challenge gives you a slow Postgres query in your browser and five minutes to fix it. If you find yourself guessing, that’s where your delta starts.

6 to 10: you make decisions, but can’t defend them yet

You make real calls at work, and the interview questions are where you slipped. Explaining a decision under pressure is what a product-company interview tests.

Play The Latency Game next. It’s seven questions in three minutes. If you can’t tell 50 µs from 5 minutes, “what breaks first at 10×?” will catch you too.

11 to 16: you’re the one deciding

Your job is safer than most, as long as the decisions keep coming. Your risk is the next project: one more year of change requests and this score drops. Take the test again in three months on a fresh set of 10 tickets and see which way it moved.

How the delta grows

You make a decision, someone who has made it before marks it up, and you redo it until you can defend it.

That’s how Vyson works. You build a backend for a million users, every submission comes back with my red pen on it, and you redo it until you can explain every choice to an interviewer, because I’ll already have made you defend each one.

If you’re just pressing Accept, Accept, Accept, you’ll be phased out. Soon. Become the engineer who decides what gets accepted.