The 3 Areas of HR AI
HR work suits AI-ification but is the area with the highest discrimination/bias risk. Careful design is needed.
1. Hiring
Document Screening
- Structure work history from resumes
- Fit score against job requirements
- Comparison with similar candidates
Caution: an architecture that excludes protected attributes (sex, age, nationality) from AI input is essential so it doesn't discriminate. Penalties exist under US EEOC and various national laws.
Interview-Question Generation
- Generate individual questions from the job description + candidate profile
- STAR-method (situation, action, result) questions
- Auto-create case-study problems
Interview Records and Summary
- Summary + strength/weakness list from interview recordings
- Integrate multiple interviewers' evaluations
- Emotion/speech analysis (caution: discrimination risk)
Improving Candidate Experience
- Earlier decision notice (automation, but human final approval)
- Creating rejection feedback (template + personalization)
- Instant response to candidate questions via AI chat
2. Evaluation
1on1 Summary
1on1 recording → summarize with AI. From manager's memory-reliance to data accumulation.
- Visualize frequent topics
- Track career wishes
- Early detection of attrition risk
Performance-Evaluation Comments
AI summarizes performance data + 360 feedback. Corrects the manager's subjective bias.
OKR Progress Tracking
AI auto-aggregates from Slack / email / project-management tools. Reduces manual-update effort.
3. Development
Individual Learning Plans
Propose individual learning resources (books, courses, internal mentors) from skill-gap analysis. Integrate with LinkedIn Learning, Udemy.
Career-Path Proposals
Present candidate positions from the person's strengths + interests + internal role set. Raises internal talent mobility.
Training-Content Generation
Personalize new-hire onboarding, compliance training with AI. Microlearning format improves learning efficiency.
Main Tools
| Product | Use |
|---|---|
| HireVue | Interview-video analysis |
| Eightfold | Hiring + development integrated |
| Lattice | Evaluation + 1on1 |
| Workday Skills Cloud | Skill management |
| Gloat | Internal talent marketplace |
| Moveworks | Employee helpdesk |
Legal Risks and Measures
Discrimination Prevention
- EU AI Act: hiring AI is the high-risk category, conformity assessment required
- US EEOC: published guidance on AI hiring tools
- NYC law: audit obligation for AI hiring tools
- Japan: labor law, Employment Security Act, APPI
Practical Measures
- Exclude protected attributes (sex, age, nationality, race, religion) from input
- AI scores are reference values; final decision by humans
- Periodic bias audit
- Transparency to candidates/employees (disclose AI use)
- Document an employee-data handling policy
Failure Patterns
- Past hiring data is biased, so AI is biased too
- Shifting responsibility to "the AI decided"
- Judging expression/voice in interview-video analysis (hotbed of discrimination)
- Analyzing candidates' social without consent (privacy violation)
- Performance monitoring worsening employee mental health
2026 Trends
- Spread of AI interviewers (HireVue, Sapia)
- Shift to skills-based hiring
- Quantifying employee engagement
- Mandatory response to EU AI Act high-risk requirements
Summary
HR AI's key is reconciling efficiency and fairness. Actively use document organizing/summarizing; a design where humans always make hiring/evaluation decisions is essential. Mindful of discrimination/privacy/transparency risks, guarantee operational quality with periodic audits.
