What should Sam do first?
Halvorsen Logistics: AI Résumé Screening
Halvorsen Logistics is a 5,000-employee U.S. freight and warehousing company that hires about 1,500 warehouse associates and drivers each year. Six months ago, the talent acquisition team adopted a vendor's AI résumé-screening tool that ranks applicants and automatically advances the top 40% to a phone screen.
Sam Rivera, a talent acquisition specialist, is preparing the quarterly hiring report. Sam notices that applicants who list graduation years before 1990 are advancing at a much lower rate than younger applicants with similar experience. Meanwhile, two hiring managers have praised the tool because it "cut screening time in half."
The vendor's contract describes the tool as "bias-tested," but Halvorsen has not conducted its own adverse impact review. The company has no written policy on using AI in employment decisions. Next week, Sam's manager, the talent acquisition manager, will present the tool's results to the vice president (VP) of operations. Halvorsen has a stated goal of building a workforce that reflects the communities it serves.
Show answer & explanation
Answer: C
Ask recruiters to hand-review a sample of the applicants the tool screened out before mentioning the pattern to anyone: Less effective. A sample review may help confirm the pattern, but doing it quietly delays telling the manager before the presentation, and an informal hand check is no substitute for a formal adverse impact analysis.
Contact the vendor to request its bias-testing documentation and wait for the vendor's findings before raising the issue internally: Less effective. The vendor's documentation is worth requesting, but waiting delays the internal escalation and leaves the possibly biased screening running.
Bring the pass-rate data to the talent acquisition manager and recommend an adverse impact analysis before the presentation: Most effective. Ensuring Impartiality & Fairness calls on HR professionals to raise possible bias promptly and push for evidence-based review. The employer remains responsible for adverse impact even when a vendor's tool causes it, and the manager needs this before presenting results.
Remove graduation years from résumés before they are uploaded to the tool so that applicants' ages are less visible to the algorithm: Less effective. It is a partial fix: other proxies such as years of experience or older job titles may still drive the pattern, and it skips analysis and leadership awareness.
When an AI screening tool shows possible adverse impact, escalate the data and analyze it promptly. Vendor assurances do not shift responsibility away from the employer. Several options could help, but only the most effective is keyed.