AI in HR Across the GCC: Are Your Talent Decisions Ready?

What needs to be true before we allow AI to influence consequential talent decisions? 

The issue is increasingly relevant in the region. The UAE’s Charter for the Development and Use of Artificial Intelligence emphasizes responsible use, transparency, human oversight, algorithmic bias, governance and accountability (United Arab Emirates Government, 2026). In Saudi Arabia, the National AI Risk Management Framework provides organizations with a structured approach to identifying, assessing, treating and monitoring AI-related risks (Saudi Data & AI Authority [SDAIA], 2026). 

For HR, these principles become particularly important when AI starts contributing to decisions about hiring, assessment, promotion, leadership, succession or development. 

At that point, AI readiness becomes decision readiness. 

The challenge is whether organizations have the measurement quality, evidence, fairness and governance foundations needed to trust the decisions AI supports. 

Automating a decision does not automatically improve it 

AI can create real value in HR. It can reduce administration, organize information, support reporting and make processes faster and more consistent. 

But there is a meaningful difference between automating an administrative process and using automated outputs as evidence in a decision about a person. 

A system that schedules an interview is different from one that ranks candidates. A tool that summarizes information is different from one whose output influences who enters a succession pipeline. 

The International Labor Organization has highlighted this challenge in its examination of AI in recruitment and people management. Its analysis focuses on three fundamental questions: whether the objective is clearly defined, whether the data are appropriate, and how the system has been programmed. The underlying concern is that AI can identify patterns in data, but it does not decide for itself what good performance, potential or leadership should mean (Berg, 2026; Berg & Johnston, 2025). 

For talent decisions, that leads to an important principle: 

AI cannot compensate for a poorly specified decision. 

If the organization has not clearly defined what it is trying to measure, automation can scale an uncertain objective. If the data do not meaningfully represent the characteristic or outcome that matters, more sophisticated analysis does not make the evidence stronger. And if the system cannot be understood, challenged or monitored, its recommendations become harder to govern responsibly. 

Before scaling AI into talent decisions, HR leaders should therefore examine four foundations. 

1. Construct validity: are we measuring the right thing?

Before asking what an AI system can predict, ask: 

What exactly are we trying to measure? 

If an AI-enabled assessment claims to evaluate cognitive ability, personality or another characteristic, the organization should be able to explain what that construct means and why it matters for the decision. 

This is fundamentally a validity question: does the evidence support interpreting the assessment as measuring the intended construct, and is that interpretation appropriate for its intended purpose? 

ISO 10667, the international standard for assessment service delivery in work and organizational settings, covers assessment used for purposes including recruitment, selection, development, appraisal, promotion and succession planning. Across its client and service-provider requirements, it recognizes principles including relevance, validity, reliability, fairness and standardization as important to assessment quality (International Organization for Standardization [ISO], 2020a, 2020b). 

SIOP makes the connection to AI directly. Its guidance states that AI-based employment assessments should be evaluated according to the same core psychometric principles applied to other employment selection procedures, including evidence that they measure job-relevant attributes and predict meaningful outcomes (Society for Industrial and Organizational Psychology [SIOP], 2023). 

A sophisticated algorithm cannot compensate for an unclear definition of what it is supposed to measure or predict. 

The first question should not be “What can this technology measure?” 

It should be “Why should this characteristic influence this talent decision?” 

2. Reliability and evidence: can we trust the measurement?

Once an organization knows what should be measured, it needs evidence that the resulting information is dependable enough for its intended use. 

AI outputs can look authoritative: a precise candidate score, an employee ranking or an instant recommendation. But precision of presentation is not quality of evidence. 

HR leaders should ask how consistently the measure performs, what evidence supports the interpretation of its scores, whether those scores relate to outcomes that matter and what limitations have been identified. 

SIOP’s guidance addresses these issues explicitly, including job relevance, score reliability and consistency, fairness, appropriate use and sufficient documentation to evaluate and validate an assessment (SIOP, 2023). 

This is also a procurement issue. In 2026, the SIOP Foundation and CHRO Association published guidance for leaders evaluating AI-based employment tools from vendors. Its focus is practical: helping organisations scrutinise vendor claims, assess supporting evidence and consider risk rather than treating the presence of AI itself as evidence of quality (SIOP Foundation & CHRO Association, 2026). 

A better buying question than “How advanced is the AI?” is: 

“What evidence gives us confidence in the talent decision this system is supporting?” 

3. Fairness and population relevance: does it work for the people being assessed?

For organisations operating across the GCC and wider MENA region, assessment quality also has a language, cultural and population dimension. 

Workforces may include Arabic and English speakers, multiple nationalities and people from varied educational and cultural backgrounds. 

That does not mean an assessment developed outside the region is automatically unsuitable. Equally, translating an assessment into Arabic does not by itself demonstrate that it works appropriately for Arabic-speaking participants. 

The International Test Commission’s guidance on translating and adapting tests makes the distinction clear: adaptation requires attention to the broader cultural context, not simply conversion from one language to another. Its guidelines also address evidence of equivalence, reliability and validity when assessments are adapted across languages and cultures (International Test Commission [ITC], 2017). 

For GCC employers, relevant questions include whether language retains the intended meaning, whether instructions and terminology function appropriately, whether score interpretations remain suitable for the target population and whether cultural or contextual factors have been considered. 

Leadership assessment provides a useful example. The question is not whether there is one uniquely “GCC” model of leadership. It is whether the behaviours defined as effective leadership genuinely reflect the role, organisational environment and business context rather than being assumed to apply universally. 

The principle is straightforward: 

Do not assume difference automatically creates bias—but do not assume equivalence without evidence. 

Arabic capability, cultural context and population relevance should be considered part of assessment quality from the beginning, not simply localisation tasks added at the end. 

4. Governance and human oversight: whoremainsaccountable? 

AI can generate an output. Responsibility for how that output is used still belongs to the organisation. 

This is why “human in the loop” is not enough on its own. 

If a recruiter can override an AI recommendation but does not understand the information behind it, oversight is limited. If a manager routinely accepts whatever a system recommends, human involvement can become ceremonial. 

Meaningful oversight requires decision-makers to understand what the technology contributes, recognise its limitations, consider conflicting evidence and have genuine authority to challenge or override a recommendation. 

That direction is consistent with regional AI governance. The UAE Charter addresses transparency, human oversight, algorithmic bias, governance and accountability (United Arab Emirates Government, 2026). Saudi Arabia’s National AI Risk Management Framework similarly places emphasis on identifying, assessing, treating and monitoring AI risk (SDAIA, 2026). 

These are broad AI-governance frameworks, not HR-specific assessment regulations. But their principles have clear relevance when technology contributes to decisions that can affect someone’s career. 

For HR leaders, governance needs to establish who approved the system, who can challenge its recommendations, what happens when AI output conflicts with other evidence and how outcomes will be monitored after deployment. 

Accountability cannot simply belong to “the algorithm”. 

Before using AI in a talent decision, ask these seven questions 

HR leaders do not need to become AI engineers or psychometricians. They do need to know what evidence to request. 

1. What exactly is the system measuring? Can the capability, construct or characteristic be clearly defined? 

1. What evidence supports the measurement? What do we know about its validity, reliability and appropriate interpretation? 

1. Is it relevant to the role or business decision? Why should this measure influence hiring, promotion, succession or development? 

1. Has performance and fairness been evaluated for the populations being assessed? Consider language, culture, accessibility and population relevance. 

1. What happens when the AI output conflicts with other evidence? Define how it will be considered alongside interviews, assessment centres, performance information and professional judgement. 

1. Who can review, challenge or override the recommendation? Human oversight should involve genuine authority. 

1. How will outcomes be monitored after deployment? Look for unexpected patterns, changes in performance and evidence that the system continues to support its intended purpose. 

These questions apply whether an organisation is buying an AI talent assessment, adding AI to an existing process or developing technology internally. 

Science should enable AI—not compete with it 

None of this is an argument against AI. 

AI and automation can improve speed, scale, consistency, administration, analysis and reporting. Assessment science provides the foundation for deciding what should be measured, how well it is being measured and what conclusions the evidence can reasonably support. 

Governance determines how that evidence enters a talent decision and who remains accountable. 

That leaves HR leaders with two different questions: 

Can we automate this? 

And: 

Should we trust the decision it supports? 

The first is primarily a technology question. The second requires sound science, relevant evidence and accountable human judgement. 

Better AI starts with better talent decisions 

AI presents significant opportunities for HR leaders across the GCC. But speed of adoption should not be confused with readiness for consequential talent decisions. 

Before technology influences who gets hired, promoted, developed or identified as future leadership talent, organisations should be able to explain what is being measured, why it matters, what evidence supports it, how fairness and population relevance have been considered, and who remains accountable. 

At Elev8, our position is simple: 

AI and automation should strengthen good assessment practice—not replace the science, evidence and human judgement that make a talent decision worth trusting. 

Before you automate a talent decision, make sure you know what evidence to ask for. 

Explore Elev8’s approach to psychometric assessment and better talent decisions. 

References 

Berg, J. (2026, May 15). The messy business of managing people at work: Is AI the solution? International Labour Organization. 

Berg, J., & Johnston, H. (2025). AI in human resource management: The limits of empiricism (ILO Working Paper No. 154). International Labour Organization.