HR Tech: AI in Hiring and HR
HR Tech
What is HR tech?
- Applies across hiring, performance, training, and workforce analytics
- Three risks: bias, explainability, and privacy
- AI assists, people decide
Adopting HR tech changes the makeup of an HR manager’s work. The time spent flipping through 240 résumés shrinks, but a new task appears: recording and explaining 'why this candidate was screened out'. What shrinks is processing time, and what grows is the burden of justification. If you leave this trade-off out of the review stage, the time you saved goes right back into handling disputes. Put the time saved and the record-keeping effort in the same table and compare them before you decide whether to adopt.
In practice, it is safer to split AI involvement into three levels: drafting support (job posting copy, first-draft interview questions), information organizing (summarizing interview recordings, aggregating completion data), and screening decisions (résumé rankings, fit scores). The first two stop before their output reaches a candidate, but the third is different. Korea’s Personal Information Protection Act lets data subjects object to, or ask for an explanation of, decisions made through fully automated processing, so using AI for screening means designing advance notice and an explanation procedure alongside it. This is where HR tech parts ways with general workflow automation. If an inventory forecasting model is wrong, it is a cost problem, but if a hiring model is wrong, it is a discrimination problem.
The most common misconception is that 'a person makes the final decision, so it is not an automated decision.' If the reviewer simply approves the ranking the AI produced, the decision is human in form but automated screening in substance, and when an objection comes in there is nothing to cite beyond 'we only looked at the top 40.' Attrition-risk prediction works the same way. The moment a prediction is handed to a line manager at the individual level, that employee becomes someone to be managed, and a week is enough for rumors to spread. Use predictions only as organization-level aggregates and never attach individual labels; that is the safe line. Also check whether people overturn AI results on a regular basis, so you avoid a rubber-stamp approval process.
John KimInstructor John Kim, author of B2B Marketing Design and CEO of Nanum Consulting
Seolyeong ParkGenerative AI and AX training specialist with 17 years of hands-on experience and 200+ lectures, helping non-technical staff put AI to work right awayThe field grew out of human resource information systems and expanded into recruiting platforms and learning management systems. As countries tighten rules on hiring-algorithm transparency and anti-discrimination, governance rather than technology is now the central issue.
Consider a hypothetical case. Company A is an auto parts manufacturer with 1,200 employees. As experienced hires grew to 240 a year, its 3-person HR team was spending 20 hours a week on résumé review alone. When it considered adopting a hiring AI, the legal team put on the brakes: "If you can’t state the reason for a rejection, we can’t adopt it."
- In the first meeting, the HR head limited AI’s role to checking whether minimum qualifications are met and ruled out fit scores and ranking altogether.
- The privacy manager added a notice of automated processing and the opt-out and explanation-request procedure to job postings, and reset the application retention period to 6 months.
- Over 3 months, running alongside human review, a recruiting assistant manager cross-checked all 240 cases one by one and found that the two disagreed on 18, most of them résumés that described experience in free-form text.
- The root cause: each job family described requirements in its own terms, so HR unified the required-qualification wording for all 12 job families into a single one-page standard.
- Six months later, first-round résumé screening dropped from 20 hours a week to 7, and the rejection-reason logging rate held at 100%, so all 4 appeals received a written reply within 3 days.