RGH

RGH-Global | Insights
AI now supports everything from shortlisting to skills screening and workforce planning. Yet as adoption increases, so does a far more difficult question, one that many organisations are only beginning to confront:

Who is responsible when recruitment AI makes a bad call?

For years, digital transformation has prioritised speed, scale and efficiency. Algorithms promised objectivity, consistency and a buffer against human bias. But the reality is more complex. AI does not remove bias, it reflects the data, design and decisions behind it. When a qualified candidate is screened out, when a particular demographic is systematically overlooked or when an algorithm produces skewed outcomes, accountability becomes unclear.

The repercussions sit squarely with real people, but responsibility often does not.

The Illusion of Neutral Technology

AI-driven recruitment systems are frequently described as “objective” or “data-led.” Yet research from MIT, Stanford and the OECD shows that algorithms trained on historic data often replicate or amplify existing inequalities. When past hiring decisions reflect certain patterns, whether intentional or systemic, the model learns them as “success signals.”

The result is not neutrality, it’s repetition.

If a CV parser downgrades candidates with career breaks, non-linear paths or certain geographical histories, is that a technical flaw or a business bias embedded in the model? And when this leads to missed talent, who is answerable?

Vendors point to user configuration. Clients point to the algorithm. HR teams point to leadership mandates. Leadership points back to process.

This loop appears only when tools operate without transparency. When platforms provide explainable outputs, as responsible AI systems do, accountability becomes shared, not avoided.

Blind Recruitment Isn’t a Safety Net

Blind recruitment has been positioned as a fairness solution, yet even this approach raises ethical questions. Removing names, universities or demographic markers may reduce certain biases, but it can also conceal systemic issues that require attention.

More importantly, blind processes do not prevent algorithmic bias in the features that remain visible. Skills clustering, work-history analysis and predicted performance modelling can all reproduce unfair patterns, even when explicit identity markers are removed.

Eliminating visible identity does not guarantee equitable outcomes. It simply hides the parts organisations may be reluctant to confront.

When AI Rejects Talent, the Cost Is Real

One of the most challenging scenarios emerges when a candidate rejected by AI later proves to be exceptional, through referral, a future application or industry reputation.

Who carries responsibility for that lost opportunity?
Was it:

• the HR team who trusted the system?
• the leadership team who approved it?
• the vendor who developed the model?
• or the data it was trained on?

This isn’t a theoretical exercise. It impacts revenue, capability, innovation, diversity and employer brand.

A misjudged automated rejection can remove talent that human recruiters would have shortlisted, interviewed and hired. As screening volume increases, the risk compounds.

The Ethical Responsibility Gap

Regulators are starting to respond. The EU AI Act, the UK AI Safety Institute and similar global bodies are emphasising transparency, testing and human oversight. Yet compliance alone does not answer the deeper organisational question:

Where should ethical responsibility sit?

Many organisations place technical accountability on the vendor, operational accountability on HR and strategic accountability on leadership. But AI-driven recruitment sits at the intersection of all three, which means no single owner is sufficient.

A credible governance framework requires:

• human-in-the-loop decision-making
• transparent audit trails
• regular bias and impact testing
• clear escalation paths for unexpected outcomes
• internal capability to understand how decisions are made

Without this structure, organisations risk delegating responsibility to a process rather than a person which creates ethical and operational exposure.

Are We Complicit in Bias If We Don’t Intervene?

This is the uncomfortable question and one that the industry must face.

If algorithms reflect the data they are trained on… If underrepresented groups face higher algorithmic rejection rates… If historic hiring data contains patterns we would no longer endorse…

Are organisations complicit if they deploy these systems without intervention?

Technology is not the issue; delegated accountability is.

Leaders cannot outsource ethical responsibility to a model; HR cannot outsource fairness to automation and vendors cannot claim neutrality when their systems shape real opportunity.

A shared-risk model is no longer optional; it’s essential.

Where Ethical AI Fits In And Why Explainability Matters

The challenge is not the presence of AI in recruitment. It is the absence of visibility into how decisions are made. Recruitment tools that operate as closed, fully automated systems create the accountability gaps organisations are now trying to close.

This is why models built around explainability, skill-based assessment and human decision-making are becoming the preferred approach.

Platforms like Epitome sit firmly within this category. Instead of making decisions on behalf of organisations, they reveal the reasoning behind matches. They show how skills align to roles, where capability gaps exist and why certain candidates surface in the process. Most importantly, they preserve human judgement at every stage.
Organisations do not need less technology, they need technology that is transparent, auditable and aligned with ethical hiring practice.

When AI supports decisions rather than replacing them, accountability becomes clearer, and hiring becomes both faster and fairer.

As AI Becomes Central to Hiring, Accountability Becomes Central to Strategy

AI will continue to evolve, and its role in recruitment will grow. Because of this, the accountability question will become more important, not less.

Leaders must treat recruitment AI with the same rigour they apply to financial modelling, governance and risk management. Transparency, explainability and oversight are now as vital as speed and efficiency.

The greatest risk in AI-driven recruitment isn’t bias, automation or complexity. It’s the belief that the system, rather than the organisation, is the one making the decision.