You Dropped Degree Requirements. Your AI Still Has a Bias Problem.
Skills-based hiring is having its moment. Seventy percent of employers now say they have implemented skills-based hiring practices, and 53% have removed degree requirements for at least some roles. The press releases write themselves: meritocracy, equity, wider talent pools.
There is just one problem. Most of these companies still funnel every applicant through AI screening tools trained on the same credential-heavy, historically biased data that degree requirements were supposed to fix. Removing the gate while keeping the gatekeeper is not reform. It is rebranding.
The PR Problem: Dropping the Gate, Keeping the Gatekeeper
When a company announces it no longer requires a four-year degree, it signals openness. But signals are not systems. According to Harvard Business Review, 90% of employers now use automated or algorithmic candidate screening in their hiring pipelines. These tools were trained on historical hiring data — data shaped by decades of credential-first filtering. As HBR puts it, AI tools trained on that data "reward candidates who look like yesterday's workforce while penalising those who deviate from legacy norms."
So the degree requirement disappears from the job posting, but the AI still down-ranks candidates without degrees, without pedigree universities, without the right sequence of corporate brand names on their résumé. The bias does not vanish. It migrates downstream, out of sight.
The Oversight Illusion: Humans Follow the Machine
The standard corporate defence is "human-in-the-loop." Eighty percent of organisations say they rely on human review as a safeguard against AI bias in hiring. It sounds reassuring. It is also false comfort.
A University of Washington study published in November 2025 tested what happens when recruiters receive AI-generated candidate rankings that contain racial bias. The results were stark: in severe-bias scenarios, recruiters mirrored the AI's biased recommendations 90% of the time. Without AI assistance, the same recruiters showed almost no racial bias in their evaluations.
As HR Dive reported, human recruiters are "perfectly willing to accept AI biases" — not out of malice, but because the tool confers authority. The Washington Post's coverage of the same research noted that even participants who were warned about potential AI bias still deferred to the machine's rankings.
The implication for CHROs is uncomfortable: the very safeguard you are citing in your compliance documentation — human review — is the safeguard that does not work. You are not catching bias. You are laundering it.
The Meritocracy Trap: Scaling Inequity
Here is where the stakes compound. Skills-based hiring is supposed to expand opportunity for candidates from non-traditional backgrounds — workers who learned through bootcamps, military service, self-study, or on-the-job experience. But when the AI screening layer is trained on historical data that correlates skills with credentials, it systematically penalises precisely those non-traditional paths.
HR Brew reported in April 2026 that the gap between the rhetoric of skills-first hiring and the reality of AI-powered screening is widening, not narrowing. The technology scales whatever patterns it learned — and if those patterns encode credential bias, racial bias, or socioeconomic bias, skills-based hiring becomes a faster, more efficient engine of inequity. You are not building a meritocracy. You are automating the illusion of one.
What an Audit-Ready Approach Actually Looks Like
The answer is not to abandon AI screening. It is to demand transparency, documented fairness testing, and genuinely auditable decision pathways.
That means knowing exactly how your screening tool scores candidates, what data it trained on, whether fairness testing has been conducted and documented, and whether every decision can be traced and explained. Tools like OVI demonstrate what this looks like in practice: transparent AI CV screening with documented fairness testing, auditable decisions, and human-in-the-loop review that functions as genuine oversight rather than rubber-stamping. Starting at $99/month, it is accessible even for mid-market teams that cannot afford enterprise compliance infrastructure.
The CHRO Checklist
Before your next skills-based hiring announcement, ask your talent acquisition team three questions:
- What data trained our AI screening tool? If the answer is "historical hiring data" with no fairness audit, you have inherited every bias your organisation ever had.
- Can we explain every rejection? If you cannot trace why a candidate was screened out, you cannot defend it — not to regulators, not to candidates, not to your board.
- Does human review actually override the AI? The UW study says it almost certainly does not, unless you have designed your process specifically to counteract automation bias.
Dropping degree requirements is a start. But if your AI is still screening for the same proxies, you have not changed the outcome. You have just changed the story you tell about it.
Sources
- HR Brew (April 2, 2026): "Well, AI Still Hasn't Solved Bias in Hiring" — https://www.hr-brew.com/stories/2026/04/02/well-ai-still-hasn-t-solved-bias-in-hiring
- Harvard Business Review (January 2026): "AI Has Made Hiring Worse, But It Can Still Help" — https://hbr.org/2026/01/ai-has-made-hiring-worse-but-it-can-still-help
- University of Washington (November 2025): "People mirror AI systems' hiring biases, study finds" — https://www.washington.edu/news/2025/11/10/people-mirror-ai-systems-hiring-biases-study-finds/
- HR Dive: "Human recruiters perfectly willing to accept AI biases" — https://www.hrdive.com/news/human-recruiters-perfectly-willing-accept-ai-biases/805585/
- Washington Post (November 25, 2025): AI hiring bias research — https://www.washingtonpost.com/business/2025/11/25/biased-ai-hiring-research-university-of-washington-study/