You are applying for a developer job. First interview: no recruiter. A voice asks you questions while you work on a real project. You are allowed to use an AI assistant. And at every step, the voice asks you: why did you make that choice?
This is not science fiction. It is Chakra, the artificial interviewer that the American platform HackerRank has just opened up to all its client companies.
How it works
HackerRank is a platform used by many companies to recruit developers. Over six months of testing, Chakra has already conducted more than 500,000 interviews. Among the companies that have tried it are the American firm Snowflake and the French group Capgemini.
The system replaces in a single interview what until now required three stages: an initial conversation with a recruiter, an exercise to do at home, then a technical interview with an engineer.
Behind the scenes, three AIs share the work. The first turns the job advert into an interview plan. The second conducts the conversation, out loud. The third scores the answers by following that plan. When a criterion could not be assessed, it is flagged as such rather than scored at random. At the end, a detailed report is sent to the recruiter, and it is a human who decides whether to hire.
What it really assesses
The big novelty is what gets measured. Until now, programming tests mainly looked at the result: does the code work? Chakra looks at the path. How the candidate approaches the problem, what decisions they make, and above all how they use AI: what they ask it, and whether they can judge what it hands back.
HackerRank's boss puts it this way: when an AI can write code that works, looking only at the result no longer says much. This is exactly the idea we were developing about exams in the age of AI: assessing the reasoning rather than the final copy.
According to the company, allowing AI during the interview has even reduced cheating: reports of suspicious activity are said to be 70 to 80% fewer than with the old tests. When the tool is allowed, there is no longer any need to hide it.
How to prepare
Think out loud. The AI assesses your reasoning. A silent candidate who finds the right answer may score lower than a candidate who clearly explains their approach.
Use AI openly, but check it. What matters is not doing without it, but showing that you can judge what it produces. Say so when you spot an error in its answer.
Justify your choices. Be ready to explain why you chose one solution rather than another. It is the question that comes up most often.
Do not cheat. The system detects unauthorised applications and help coming from a second device. It is not worth the risk.
Practise under the same conditions. Solve a small problem with an AI assistant, explaining each step out loud. It is an unusual exercise, and it can be learned.
In Europe, an AI used for recruitment is classified as "high risk" under the EU AI Act, which notably requires human oversight and transparency. And the data protection regulation, the GDPR, gives you the right not to be subject to a decision based solely on automated processing that significantly affects you. You can request human intervention, express your point of view and contest the decision. Do not hesitate to ask the recruiter who will read the report, and how the decision will be made.
What we take away
For now, this system mainly concerns technical roles, even though HackerRank claims it can also be used for other positions. And it will not arrive everywhere overnight.
But it shows a direction. As the French study on job adverts reminded us, what is expected of candidates is changing. Knowing how to do the job matters less. Knowing how to explain, judge and decide matters more and more. Including in front of a machine.