AI Is Required at Work, But Banned in the Interview
The tech industry has a strange rule: use AI every day at work, but pretend it does not exist when you are trying to get the job.
There is something funny about the current state of tech hiring.
Many companies want employees who know how to use AI.
They mention AI tools in job descriptions.
They expect engineers to use them at work.
They want people who can work faster with AI.
But then the interview starts.
Suddenly, AI is not allowed.
No ChatGPT.
No coding assistant.
No AI-powered search.
Sometimes, not even autocomplete.
This creates an obvious question:
If AI is part of the job, why should candidates pretend they do not know how to use it?
That is the debate happening across the tech industry.
The irony is hard to miss
A recent post by shafu captured the situation in a simple sentence:
The post struck a nerve.
Thousands of people reacted.
Many engineers shared similar experiences.
The problem is not really about AI alone.
It is about what companies are trying to measure.
If a company wants to know whether someone can write code without help, then an AI-free interview makes sense.
But if the actual job involves using AI every day, the test may not represent the real job.
That is where things get confusing.
The workplace has already changed
AI coding tools are now part of many developers’ workflows.
A developer may ask AI to explain an error.
They may use it to write a first draft.
They may ask it to create tests.
They may use it to review code.
They may also use AI to understand an unfamiliar library.
None of this automatically means the developer lacks technical skills.
The skill is changing.
It is becoming less about typing every line manually.
It is more about knowing what to build.
It is about asking the right questions.
It is about checking the output.
It is about finding mistakes.
And it is about knowing when not to trust AI.
That is a very different skill from traditional coding.
But companies have a valid concern
There is another side to this debate.
Companies still need to know whether a candidate understands the basics.
Imagine hiring an engineer who cannot explain their own code.
That would be a problem.
AI can produce impressive code.
But impressive code does not always mean correct code.
It can contain security problems.
It can use the wrong approach.
It can miss important edge cases.
It can also confidently give an incorrect answer.
So companies have a reason to test independent thinking.
The real question is whether banning AI is the best way to do that.
The calculator comparison
This debate is not completely new.
Remember calculators?
Schools once debated whether students should be allowed to use them.
If a math test was designed to test basic arithmetic, a calculator could defeat the purpose.
But in real life, nobody cares if an engineer can multiply large numbers in their head.
The calculator is a tool.
What matters is understanding the problem.
AI may be heading in the same direction.
In the future, knowing how to use AI could become as normal as knowing how to use a search engine.
The important skill may not be producing the answer manually.
It may be knowing whether the answer is actually good.
Not every interview needs AI
This does not mean every interview should allow AI.
There are situations where banning it is reasonable.
A junior developer may need to demonstrate basic programming knowledge.
A company may want to understand how someone thinks without external help.
A debugging exercise may be designed specifically to test raw problem-solving skills.
In these cases, an AI-free test can be useful.
But companies should be honest about what they are testing.
If the test is supposed to measure independent coding ability, say so.
Do not call it an AI-friendly workplace and then secretly treat AI use as cheating.
The opposite approach is also possible
Some companies are taking a different path.
Instead of banning AI, they can make AI part of the interview.
Give the candidate a real problem.
Allow them to use AI.
Then ask them to explain their decisions.
Ask why they accepted one suggestion and rejected another.
Ask them to find bugs in AI-generated code.
Ask them to improve the result.
This could provide a much better picture of modern engineering skills.
After all, this is closer to how many developers actually work.
What if AI does most of the work?
This is another valid concern.
Suppose a candidate gives a problem to an AI tool.
The AI produces almost everything.
The candidate simply copies the answer.
What exactly has the company tested?
Probably not much.
That is why companies need better interview design.
Instead of simply asking candidates to solve a problem, interviewers can ask them to explain their approach.
They can change the requirements halfway through.
They can introduce a bug.
They can ask the candidate to modify the solution.
Someone who understands the work should be able to adapt.
Someone who only copied an AI answer may struggle.
AI bans can also create an unfair signal
There is another issue.
Not everyone uses AI in the same way.
Some developers use it heavily.
Others use it only for small tasks.
Some companies provide expensive AI tools to employees.
Others do not.
Candidates are therefore entering interviews with very different levels of AI experience.
A strict ban may not always reveal who will perform better at the actual job.
It may simply reward people who are good at traditional interview exercises.
That is not necessarily the same thing.
Job descriptions need to catch up
Companies also need to decide what they actually want.
If a job description says:
“Strong AI skills required.”
Then the interview should probably test those skills.
If AI is important to the role, candidates should have a chance to demonstrate how they use it.
Otherwise, there is a disconnect.
The company is essentially saying:
“We want you to use this tool after joining us, but we don’t want to see you use it before joining us.”
That is a strange hiring strategy.
The future may be somewhere in the middle
The answer is probably not “AI everywhere.”
It is also probably not “AI nowhere.”
A better approach could be to have different interview stages.
One stage could test basic skills without AI.
Another could allow AI.
A final stage could test real-world problem solving with AI.
This would give companies a better understanding of the candidate.
It would also reflect how modern technical work actually happens.
The real skill is becoming judgment
AI is changing what it means to be good at technology.
Knowing how to code still matters.
But knowing what to code matters more.
Knowing how to check AI output matters.
Knowing how to spot bad assumptions matters.
Knowing how to ask better questions matters.
Knowing when to use AI matters too.
And sometimes, knowing when not to use it matters most.
That is probably the skill companies should be testing.
The tech industry is still figuring this out.
Some companies will ban AI.
Some will allow it.
Some will require it.
The rules will probably continue to change.
But one thing is already clear.
The interview should reflect the job.
If AI is a normal part of the job, pretending it does not exist during the interview may not make much sense.
The better question is not:
“Can you work without AI?”
It is:
“Can you do good work with AI?”
That is the skill the next generation of tech hiring may need to measure.
—
This post was previously published on Sushila Devi’s blog.
***
If you believe in the work we are doing here at The Good Men Project, please join us as a Premium Member today.
All Premium Members get to view The Good Men Project with NO ADS.
Need more info? A complete list of benefits is here.
Photo credit: iStock
