AIWiki
Malaysia
Back to all articles
Applicationshuman resourcesrecruitmentautomated decision-making

AI in Human Resources

4 min readUpdated November 2026
AI in Human Resources
Type
Industry application
Main tasks
Screening, matching, onboarding, retention, employee support
Typical systems
Applicant tracking systems, HR copilots, workforce analytics
Key risks
Bias, opacity, unlawful automated rejection, surveillance
Key regulation
PDPA Automated Decision-Making and Profiling Guideline (2026)
Related
AI bias, AI and employment, AI agents
AI in human resources is the application of machine learning and language models to people processes: sourcing and screening candidates, ranking résumés, answering employee questions, analysing skills, forecasting attrition and administering onboarding. It is one of the most consequential and most regulated deployments of AI, because its outputs decide who gets hired, promoted or exits — decisions with direct legal and financial effects on individuals.[3]

History and Background

Recruitment was automated before it was "AI": applicant tracking systems (ATS) of the 2000s filtered candidates by keywords, a practice that pushed applicants toward résumé keyword optimisation. Machine learning added scoring and ranking in the 2010s, and the 2023 wave of large language models changed the interface — screening summaries, job-description generation, conversational candidate screening and interview transcription. Regulators responded in parallel: the European Union's GDPR had already constrained solely automated decisions with legal effect under Article 22, and Malaysia's Personal Data Protection Commissioner published its Automated Decision-Making and Profiling (ADMP) Guideline on 30 April 2026, alongside companion guidelines on data protection impact assessments and data protection by design.[1][2]

Key Concepts and Technology

The recurring components are: sourcing (semantic search over candidate pools and job boards), screening and ranking (models scoring résumés against a role, often using embeddings rather than keywords), conversational screening (an agent asking structured questions and summarising answers), interview analysis (transcription, competency tagging and — controversially — sentiment or engagement scoring), workforce analytics (skills graphs, attrition and flight-risk prediction, workforce planning) and employee copilots (answering policy, leave and payroll questions from internal documents using retrieval-augmented generation).

The risks are well documented. Models trained on historical hiring data reproduce historical preference — proxy discrimination can persist even when protected attributes are removed, since variables such as school, address or career gap can act as substitutes. Attribute inference, where sensitive traits are reconstructed from non-sensitive ones, is a specific concern regulators have highlighted. Jurisdictions converge on three controls: transparency (tell the person that automation is being used and why), meaningful human review (a trained human must be able to change the outcome, and merely pressing a button on system output does not qualify), and contestability (the ability to challenge or withdraw).[1][3]

Applications and Impact

Used carefully, the tools shorten time-to-hire, reduce administrative load on recruiters, improve candidate response times and make internal mobility visible. Used carelessly, they generate confident but false résumé summaries, rank candidates for reasons no human can articulate, and quietly narrow a shortlist in ways that are only discovered at audit. The emerging consensus among practitioners is narrow automation — machines prepare, compare and draft; people decide and document.

>See Also

🇲🇾Malaysian Context

🇲🇾 Relevance to Malaysia: the compliance position changed materially on 30 April 2026, when the Personal Data Protection Commissioner under the Ministry of Digital issued the ADMP Guideline, the DPIA Guideline and the Data Protection by Design Guideline under the Personal Data Protection Act 2010.[1][2] Two features matter for HR. First, automated decision-making is defined as decisions made without meaningful human involvement — a process stays in scope even where a person merely inputs the data and the system decides; the guideline's own example is a human-resources department using an algorithm to rank and shortlist candidates for interview. Second, ADMP is a qualitative trigger: any use of automated decision-making or profiling requires a data protection impact assessment regardless of volume, separate from the quantitative thresholds of 20,000 data subjects (or 10,000 for sensitive personal data).[2] Practically, a Malaysian employer adopting an AI screening tool must be able to inform applicants that automation is used, explain the reasons and consequences, keep a trained reviewer in the loop, honour withdrawal of consent, and document the assessment before the system goes live. These obligations sit alongside the broader national agenda: the National AI Office and Malaysia's talent-development programmes frame AI literacy for HR and hiring managers as workforce infrastructure, while HRD Corp-funded training channels support upskilling for the same staff who will operate, audit and govern these systems.

References

  1. ↑Personal Data Protection Commissioner, Malaysia (2026). Automated Decision-Making and Profiling (ADMP) Guideline. https://pdp.gov.my/wp-content/uploads/2026/04/Automated-Decision-Making-And-Profiling-Guideline-ADMP.pdf
  2. ↑InCorp Malaysia (2026). DPIA Malaysia rules for AI tools and automated decision-making. https://malaysia.incorp.asia/blogs/dpia-malaysia-automated-decision-making
  3. ↑Edwin Lee & Partners. A full guide to automated decision-making and profiling (ADMP) in Malaysia. https://lpplaw.my/insights/e-articles/admp-malaysia
  4. ↑HELP University (2026). The use of artificial intelligence in the recruitment process. https://university.help.edu.my/wp-content/uploads/2026/02/12-WAN-ARIFAH-MEOR-KAMAL-AZHAR-THE-USE-OF-ARTIFICIAL-INTELLIGENCE-IN-THE-RECRUITMENT-PROCESS-23FEB26-1.pdf
Back to all articles