Unilever case Study | HR case Study | Unilever Recruitment | AI in HR | Artificial Intelligence

Overview

• Case study of Unilever’s experiment using AI-driven assessments to scale hiring for its Future Leaders program (~50 countries).
• Pilot combined game-based assessments (pemetrics/pemetrics) and on-demand video-interview AI (higher view/HireVue) to automate screening and predict job performance.
• Business goals: attract millennial talent (target cited: 60% of workforce millennials by 2020), speed hiring, reduce costs, and improve diversity and candidate experience.

Process / playbook (operational flow)

Replace paper/phone/manual screening with a digital funnel:

1. Online application
2. Game-based cognitive & behavioral assessment (pemetrics) to measure traits such as aptitude, reasoning, risk appetite, and motivational drive
3. On-demand asynchronous video interview (higher view) — candidates record responses
4. AI analysis: extract features from video and games, produce an insight/ranking score
5. Shortlist candidates for assessment centers where human recruiters conduct final selection

• Combined human + AI workflow: automated pre-screening at scale, human-led assessment centers for final hiring decisions.
• Pilot → measure → scale approach across a global leadership intake.

Frameworks / techniques highlighted

• Digital hiring funnel / hybrid automation playbook (AI screening + human confirmation)
• Game-based assessment as behavioral/psychometric measurement (adaptive, progressive)
• Asynchronous video interviewing + ML scoring (cited: ~25,000 signal features)
• Objective-data-first selection to reduce subjective bias (resume/interpreter bias reduction)
• Pilot → measure → scale for iterative rollout

Key metrics, KPIs, targets, timelines

• Applicant volume: 250,000 applicants initially
• Shortlisted to assessment centers: 3,500
• Final hires: 800

Time / cost

• Saved ~50,000 candidate interview hours over 18 months
• Reported ~£1 million annual cost savings
• Historically took 4–6 months to screen 250,000 applications; AI reduced screening time (specific baseline not fully quantified)
• Reported 90% reduction in time-to-hire

Candidate engagement / quality

• 96% candidate completion rate (vs 50% baseline)
• 16% increase in diversity (reported)

AI scoring / features

• Example insight score: 65% for one candidate
• Feature set: ~25,000 features analyzed (facial expressions, eye contact, body movement, voice nuances, clothing details)

Concrete examples / case evidence

• Media demo: Business Insider reporter Rich Filoni (Rich Feloni in some citations) took the on-demand interview, received a 65% insight score and was ranked second in that pool.
• Outcome for Unilever’s Future Leaders program: reduced applicant pool from 250k → 3.5k invited to assessment centers → hired 800; reported savings and diversity gains as above.

Example insight: one candidate was given a 65% “insight score” and ranked second in their pool.

Benefits

Candidate benefits

• More engaging assessments via games versus traditional psychometrics
• Flexibility through asynchronous interviews (time-zone coverage)
• Immediate feedback from game levels; potentially less stress and stereotype-threat
• Perceived reduction in human bias due to objective data

Client benefits

• Faster, scalable screening of very large applicant volumes
• Richer data beyond resumes (behavioral and nonverbal cues)
• Reusable recorded interviews for review and calibration
• Cost and time savings; improved completion rates and diversity

Actionable recommendations (transferable tactics)

• Use game-based assessments to measure cognitive and behavioral traits that resumes miss; ensure tests are adaptive and emphasize process over “right answers.”
• Deploy asynchronous video interviewing for scale and global time-zone coverage; analyze video with ML to surface nonverbal and vocal predictors.
• Combine AI pre-screening with human-led assessment centers for final hires to retain human judgment for high-stakes decisions.
• Pilot with a targeted program (e.g., leadership intake), measure completion rates, time-to-hire, cost per hire, and diversity outcomes before scaling.
• Track and publish metrics: candidate completion rate, time-to-hire reduction, hours saved, cost savings, diversity lift, and predictive validity against job performance.
• Validate ML models against actual performance outcomes to avoid false positives and ensure fairness.

Risks, caveats, and governance considerations

• AI recruiting is “still in its infancy” — requires careful validation and continuous monitoring for bias and adverse impact.
• Transparency: insights and scoring should be explainable to candidates and regulators.
• Privacy and consent: recording and analyzing video and behavioral games raises data-protection requirements (e.g., GDPR).
• Human oversight remains important to catch model errors and preserve candidate experience.

Presenters / sources

• Harry — speech robot narrator used in the video
• Rich Filoni (Rich Feloni in many citations) — senior strategy reporter, Business Insider (appears in demo)
• Lauren Larson — cited as Chief Technology Officer at “higher view” in the video
• Vendors referenced: “pemetrics” (game-based assessment) and “higher view” (video interviewing)
  • Note: subtitles contain possible misspellings; likely referring to Pymetrics (game-based) and HireVue (video interviewing)
• Video creator/channel: 5 Minutes Learning (YouTube)