Interviews that let you use AI are SCARIER than you think

Kelvin Graddick · 3 minute read ·

If they’re letting you use AI in a technical interview, it might end up being harder than you think.

At first glance, this sounds like a win. AI assistance? Autocomplete? Faster coding? Seems like that should make the interview easier, right? Not necessarily.

Based on perspective, it could actually be harder than traditional coding challenges or similar task-based interviews.

Why AI-friendly interviews can be more difficult

I was exaggerating with the “10 minutes” example, but you get my point.
If companies allow AI, they’re almost guaranteed to add another major constraint.

They might ask you to:

Build a fully functional app in 20–30 minutes
Design something end-to-end, not just a single algorithm
Make architectural decisions on the fly
Explain why your solution works, not just that it runs

You might hear something like:

“You’ve got AI, so if you really understand software development, you should be able to build an app in 30 minutes or less.”

Or even:

“We want you to build a working app and put it on the App Store for beta testing in 20–30 minutes.”

Because in their mind:

You have AI. You should be way faster. You should make less mistakes.

AI doesn’t lower the bar, it moves it

Letting candidates use AI does not mean the interview gets easier.
It means the expectations shift.

Instead of testing:

Syntax recall
LeetCode-style memorization
Isolated algorithms

They’re testing:

Problem decomposition
System design instincts
Product thinking
Speed of decision-making
Your ability to guide AI effectively

AI removes the excuse of “I forgot the syntax.”
What’s left is whether you actually understand what you’re building.

And that’s a much harder thing to fake.

Why these interviews can be better (but not easier)

I don’t think companies are going to make interviews easier just because AI is allowed.

In fact, I think the opposite happens.

But… I do think this can lead to better interviews.

Traditional coding interviews often reward:

Memorization
Pattern recognition
Practicing the same problems over and over

AI-assisted interviews can reward:

Real-world engineering skills
Judgment
Tradeoff analysis
Communication
Speed under ambiguity

That’s much closer to what the actual job looks like.

So what should candidates do?

If AI is going to be allowed, you need to train differently.

1. Use AI to learn fundamentals, not just generate code

Make AI teach you the concepts behind the domain you’re interviewing for.

Examples:

If it generates an API, ask why that structure was chosen
If it suggests a database schema, ask about tradeoffs
If it writes a component, ask how state and lifecycle are managed

Don’t blindly accept output.
If you can’t explain what the AI wrote, that will show immediately.

A good starting point:

2. Learn how to actually build with AI under pressure

There’s a huge difference between:

Using AI casually
Using AI efficiently under time constraints

You should be familiar with:

How to prompt clearly
When AI is likely to hallucinate
When to override or simplify its suggestions
How to scaffold fast and refine later

AI is not magic.
It’s a multiplier and multipliers amplify both skill and confusion.

Some good reads:

The new "AI" coding interview

This new type of coding interview isn’t:

“Can you code without help?”

It’s instead:

“Can you build real things quickly, explain your decisions, and use AI as a tool instead of a crutch?”

That’s a much higher bar.

Final thoughts

AI-friendly interviews probably won’t be easier.
But they might be fairer, more realistic, and more aligned with real work.

And honestly, that’s not a bad thing.

What’s your take on this?

Want to share this?