Knowledge · AI Basics

AI Interview · AI Competency Assessment

AI Interview and AI-based Competency Assessment

📂AI Basics⏱2 min read📊Advanced🔗6 related

What is an AI interview?

In one line
A selection tool in which AI analyzes candidates' video answers or online task records and scores their competencies and job fit. You can use it only if you can explain what it measures and why.
Bring this topic in-house📚 Related Course💬 Ask About Training
⚡ Summary in 3 lines
  • A selection tool that turns inputs into scores
  • Review three things: validity, fairness, and explainability
  • AI provides reference material; people decide and keep the record

Treating AI interviews as 'technology that replaces interviewers' is risky. The tool turns inputs, such as a candidate's answers and task records, into scores. Whether those scores actually relate to performance after hiring has to be verified separately. A manager evaluating adoption should ask the vendor not for an accuracy figure but 'what does it measure, and what is the evidence?'

There are three review criteria. The first is validity: does the competency being measured connect to the requirements of the role, and is there evidence validating the relationship between scores and the performance of current employees? The second is fairness: check whether scores differ between groups by gender, age, speech style, disability, or recording environment. Scoring non-verbal signals such as facial expression or voice has long drawn concerns about its evidence base and its potential for discrimination, so you need to confirm what actually feeds into the score. The third is explainability: when a rejected applicant asks, you must be able to explain what was evaluated.

Laws and guidelines also matter. Korea's Personal Information Protection Act gives data subjects the right to request an explanation of, or to refuse, a fully automated decision, and how it applies may depend on whether AI results alone decide who passes or people use them only as reference material. The high-impact AI discussion under the AI Basic Act also mentions decisions that significantly affect individual rights, such as hiring. There has also been ongoing discussion of adding notice of AI use to the Fair Hiring Procedure Act. Details may change, so check the latest guidance from the relevant agencies, such as the Ministry of Employment and Labor and the Personal Information Protection Commission.

The common mistake in practice is using an AI score as a blanket cutoff that rejects everyone below a set threshold. The more applicants you have, the stronger the temptation, but screening people with an unvalidated score means you miss qualified candidates and take on dispute risk. Failing to tell applicants that AI is used and what it evaluates is also a problem. The safer structure treats AI output as reference material for human judgment, with a person recording the final decision and the reasons for it.

Courses that cover this term · AI & DX
Recommended courses · Recommended instructors · Quick quote
📚AI Training Pre-Assessment and Customized Training DesignOnline assessment + 2 hours · In-person corporate training · John KimView course →📚Vibe Coding Security: Safety-Checking Your Own AI Web Apps4 hours · In-person corporate training · John KimView course →
Instructor John KimJohn KimInstructor John Kim, author of "B2B Marketing Design" and CEO of Nanum ConsultingInstructor Seungtae ParkSeungtae ParkAX instructor who teaches by building AI business development and marketing automation himself
🧮 Mini quote · Based on instructor fees
About KRW 1.2 million (≈ US$900)2-hour talk · up to 20 people · Korean
Get a quote for these requirements →
💡 In plain English
A scale can show a number accurately, but you can't judge running ability from that number. In the same way, whether a score connects to job performance has to be verified separately.
Why it matters in practice
You may be considering it as a way to lighten résumé screening and first-round interviews while setting next year's hiring volume and process, but fairness disputes and legal obligations mean you need clear decision criteria.
Origin and sources

The term came into use as selection tools that combine video interviews and online assessments with automated analysis became widespread; it has no specific originator. Data subjects' rights regarding automated decisions are set out in Article 37-2 of the Personal Information Protection Act, and the scope of application should be confirmed through guidance from the relevant authority.

Source · Personal Information Protection Act, Article 37-2 (Rights of Data Subjects regarding Automated Decisions)
An example

Consider a hypothetical case. Company J, a household goods distributor with 800 employees, received 2,400 applications for its second-half entry-level hiring. A four-person recruiting team could not handle first-round interviews, so it adopted an AI competency assessment and automatically rejected the bottom 60%. Three rejected candidates then asked for the basis of their evaluations.

  1. The hiring team lead asked the vendor, "What does it measure, and what goes into the score?" Only then did the team learn that, besides the answers and task performance records, facial expression and voice analysis were part of the score.
  2. The HR team had 120 second-year employees take the same assessment and compared their scores with their performance reviews. Task performance scores showed a relationship, but the facial expression and voice items did not.
  3. The legal officer concluded that automatic rejection could count as an automated decision. After checking the relevant regulator's guidance, the officer recommended: "Let's change this to a structure where people make the decision."
  4. Starting with the next hiring round, we turned off the facial-expression and voice items and limited AI scores to reference material for interviewers. Job postings and the test screen now state that AI is used, what it evaluates, and who to contact with questions.
  5. Two recruiters re-reviewed the bottom 30% alongside their application documents, and 41 of those candidates moved on to interviews. Of the 60 final hires, 7 came from that group.
When to use it, and when to avoid it
✅ Good fit when
  • When there are so many applicants that materials must be pre-sorted before people review them
  • When you can validate the relationship between scores and current employees’ performance in-house
  • When you are ready to tell candidates that AI is used and what it evaluates
🚫 Use with caution when
  • When you plan to accept or reject candidates automatically based on AI scores alone
  • When the vendor can't explain what goes into the score
  • When you have no process for answering a rejected candidate's request for an explanation
Related terms
HR Tech · AI in Hiring and HRThe field of applying AI and data technology to HR work such as hiring, evaluation, training, and workforce analytics. Fairness and legal requirements come with the efficiency gains.Structured Interview · Behavioral InterviewAn interview method in which every candidate is asked the same questions in the same order and evaluated on a set scale. The behavioral interview, which asks about past behavior, is the best-known example.Fair Hiring Procedure Act · Blind HiringThe law that sets employers' obligations and prohibitions to ensure fairness in hiring, and the hiring approach of evaluating candidates with personal information unrelated to the job hidden.AI BiasThe phenomenon in which AI produces unfair results for certain groups or perspectives because of skewed training data or design.Explainable AI · XAIExplainable AI (XAI) makes it possible for people to understand why an AI reached a given conclusion. It is central to trust, accountability, and regulatory compliance.AI Basic Act · EU AI ActNational-level frameworks governing the development and use of AI. The best known are Korea's Framework Act on the Development of Artificial Intelligence and Establishment of a Trust Foundation and the EU (European Union) AI Act. Both impose obligations that vary by risk level.
Training linked to this term
📚Interview Strategy, Presentation InterviewsNanum Consulting corporate training · In person or onlineView course →
Need in-house training on this topic?
Get a curriculum and quote tailored to your needs right away.
Ask About Training
📚 See Related Course🗂 All Programs📄 Free Resources🔎 Glossary Home
Last updated 14 Oct 2026 · Reviewed by John Kim

FAQ

Yes, it is safer to tell them. Rules on personal data processing and automated decisions may apply, and legislative discussions on making notice mandatory have continued. Check the latest guidance from the Ministry of Employment and Labor and the Personal Information Protection Commission for the specific requirements that currently apply.
We don’t recommend it. Deciding who passes based solely on AI output, without human involvement, may count as an automated decision, and if the scores are unvalidated, the risk of missing suitable candidates is high. The safer structure is one where people use the results as reference material and record the reasons for their decisions.
Get in writing the definition of each competency measured, the input items that feed into the score, evidence validating the relationship between scores and job performance, results of checks for score differences between groups, and the extent of explanation available to rejected candidates. Features vary by product and version.
Using them together is the common approach. The AI competency assessment produces reference material at an early stage, and human interviewers then verify it in a structured interview tailored to job requirements. Both stages should rely on the same competency definitions so that their results can be compared.