Blog 07 - Diversity and Inclusion: Can Technology Create a Fairer Workplace?

Diversity and Inclusion: Can Technology Create a Fairer Workplace?

Equality, diversity and inclusion (EDI) have become central issues in contemporary Human Resource Management. Organisations increasingly operate with employees from different genders, generations, nationalities, cultures, abilities, educational backgrounds and life experiences. At the same time, Artificial Intelligence (AI) is becoming embedded in recruitment, performance management, learning, workforce analytics and everyday work.

This combination creates an important question. Can technology help organisations make employment decisions more fairly and remove barriers for underrepresented employees, or can it reproduce existing inequality in a faster and less visible form?

AI is not automatically inclusive or discriminatory. Its effect depends on the data, design choices, organisational culture and human decisions surrounding it.

Diversity Is Not the Same as Inclusion

The distinction between diversity and inclusion is important. Diversity concerns differences within a workforce. Inclusion concerns whether those differences are respected, supported and able to contribute fully.

The CIPD (2026) argues that effective EDI should be embedded into organisational priorities, people processes, leadership behaviours and governance structures. Its purpose is not simply representation. Inclusive workplaces aim to create environments where individuals can feel safe, experience belonging and perform to the best of their abilities.

This distinction matters for technology. An organisation could use an AI system that helps recruit a more diverse group of employees while still having an organisational culture in which those employees do not feel heard, respected or able to progress. Technology might improve access, but it cannot create inclusion by itself.

AI Does Not Affect Every Group Equally

The impact of AI on employment is already uneven. The International Labour Organization’s refined global index of generative-AI exposure found that 4.7% of women’s employment falls within the highest exposure category compared with 2.4% of men’s employment (Gmyrek et al., 2025).

Figure 1. Gender differences in the highest category of generative-AI occupational exposure. Author’s illustration based on ILO (2025).

The ILO explains that this difference is strongly related to occupational structure. Women are overrepresented in clerical and administrative occupations, which contain many tasks that generative AI can potentially support or automate. The effect is even stronger in high-income economies.

However, exposure should not automatically be interpreted as job loss. The same ILO research also identifies opportunities for augmentation. The critical HRM issue is therefore whether affected employees receive access to training, redesigned roles and new opportunities, or whether technological change reinforces existing labour-market inequalities.

An equality issue can emerge even without intentional discrimination.

If one demographic group is concentrated in jobs that are more exposed to automation and receives less access to retraining, technological change can widen inequality even when the technology itself never uses gender as an input.

How Algorithms Can Reproduce Historical Bias

AI systems learn patterns from data. In HRM, those data may include previous recruitment decisions, performance ratings, promotion records or workforce characteristics. Historical data can therefore contain the effects of earlier human decisions.

The OECD (2025) warns that algorithms can either codify bias or help reduce it. Bias can enter at the data level through historical, incomplete or unrepresentative information and at the system level through decisions about which variables are selected, how they are measured and what outcomes the model is asked to optimise.

This means removing obviously sensitive variables does not necessarily solve the problem. A system may use other variables that indirectly reflect gender, disability, ethnicity, age or socio-economic background. If historical success inside the organisation was shaped by unequal opportunities, an AI system trained to identify people who resemble historically successful employees may reproduce the same pattern.

Automation does not remove bias when the algorithm simply learns yesterday’s inequality and applies it to tomorrow’s decisions.

But AI Can Also Be Used to Reduce Bias

The debate should not be one-sided. Human decision-making is also vulnerable to stereotypes, inconsistent judgement and unconscious bias. A carefully designed system can sometimes identify patterns of inequality that managers fail to notice.

The OECD (2025) provides an example from LinkedIn, which introduced an AI feature intended to make the gender composition of recruiter search results more representative of the underlying candidate pool. The same OECD analysis discusses techniques such as bias detection, balanced recommendations and synthetic profiles designed to improve representation.

This demonstrates an important point: algorithms can be designed to amplify historical patterns or to challenge them. The difference lies in the goals and safeguards chosen by the organisation.

Technology Can Expand Accessibility

One of the strongest arguments for inclusive technology concerns disability and accessibility.

The OECD’s 2025 analysis identifies 142 AI-powered solutions that can support people with disabilities in the labour market, based on research involving more than 70 stakeholders. Approximately 75% of those solutions would not exist in their current form without AI.

Figure 2. AI-powered workplace accessibility solutions. Author’s illustration based on OECD (2025).

Examples include live captioning for deaf and hard-of-hearing employees, image-recognition tools that describe surroundings to blind and low-vision users, speech-to-text applications, smart mobility technologies, AI-enhanced prosthetics and natural-language tools that can convert complex information into clearer language.

This is a different way of thinking about inclusion. Instead of requiring an employee to adapt to an inaccessible workplace, technology can help redesign the workplace around a wider range of human needs.

However, the OECD also identifies barriers: limited funding, employer awareness, affordability and insufficient user involvement in the design process. A technically impressive accessibility tool has little inclusive value if the employees who need it cannot access it.

Automated Systems Can Also Create Disability Discrimination

The same technology that can improve accessibility can also create barriers if systems are not designed for different users.

The U.S. Equal Employment Opportunity Commission (EEOC, 2024) explains that AI may be used to screen CVs, evaluate recorded video interviews, monitor employees, analyse customer feedback, influence pay and promotion decisions and even contribute to termination decisions. Existing anti-discrimination protections still apply when employers use automated technology.

For example, an automated assessment may disadvantage a candidate with a disability if it assumes that one particular style of speech, movement, eye contact or interaction represents competence. An employer may also need to provide a reasonable accommodation or an alternative assessment method rather than forcing every candidate through the same technological process.

Intersectionality Makes Inclusion More Complex

People do not experience the workplace through only one characteristic. A woman with a disability may face different barriers from those experienced by women generally or by disabled employees generally. An older migrant worker may experience technology differently from a younger employee born and educated in the same country as the organisation.

The CIPD’s 2026 EDI guidance explicitly highlights intersectionality. From an HRM perspective, this means that fairness cannot always be evaluated by looking at one demographic variable at a time. A system can appear fair at an overall level while producing unfair outcomes for a smaller group at the intersection of several characteristics.

Inclusion by Design

A responsible organisation should therefore consider inclusion throughout the life cycle of an AI system rather than testing fairness only after problems appear.

Figure 3. Inclusion-by-design framework for workplace AI. Author’s illustration informed by CIPD, ILO, OECD and EEOC guidance.

An inclusion-by-design approach should include:

• Representation: use relevant and sufficiently diverse data rather than assuming past workforce patterns represent an ideal future workforce.
• Accessibility: test systems with people who have different abilities, communication styles and access needs.
• Bias testing: examine whether outcomes differ systematically between groups and investigate why.
• Human oversight: ensure significant employment decisions remain reviewable by accountable people.
• Employee voice: involve affected workers and representatives in the introduction and review of workplace technology.
• Continuous monitoring: inclusion is not a one-time audit; organisations should monitor real-world outcomes after implementation.

Employee Voice and Social Dialogue

The ILO’s 2025 global case studies on AI and algorithmic management show the importance of social dialogue when technology affects employment and working conditions. Worker representatives across different regions have influenced decisions about algorithmic management, employment and AI implementation, helping push organisations towards more equitable approaches.

This matters because employees may identify exclusion risks that developers and senior managers overlook. People who experience a barrier directly often have the clearest understanding of why a system is inaccessible or unfair.

Inclusive technology should not only be designed for employees. It should be designed with employees.

A Global HRM Perspective

Diversity and inclusion become even more complex in multinational organisations. Brewster et al. (2017) and Varma and Budhwar (2014) emphasise that International HRM operates across different cultures, labour markets, legal systems and social expectations.

A global organisation may use a single AI platform across multiple countries, but the data used to train the system may come mainly from one region or language. Qualifications, names, communication styles and career pathways can differ considerably between countries. A system that performs reasonably well in one environment may therefore produce weaker or less fair outcomes elsewhere.

The OECD’s 2025 analysis of Korea also highlights that AI adoption itself is unequal. Larger organisations generally have greater access to digital infrastructure, skills and resources than smaller firms. This creates another inclusion issue: employees’ opportunities to benefit from AI can depend not only on who they are, but also on where and for whom they work.

Technology Cannot Replace Inclusive Leadership

An organisation can purchase bias-detection software, accessibility tools and workforce analytics. None of those technologies can substitute for leadership behaviour.

The CIPD (2026) argues that EDI needs to be connected with organisational priorities, leadership, culture and governance. Its practical guidance recommends involving employees, developing people-manager capability, securing senior commitment, evaluating people policies and examining organisational culture.

This is particularly important because managers ultimately influence whether employees receive opportunities, whether concerns are taken seriously and whether technology is used to support people or simply control them.

🎥 Recommended Video

CIPD – Inspiring Change: Championing Inclusive Leadership

This CIPD discussion explores how leadership behaviours and organisational action contribute to genuine workplace inclusion rather than treating diversity as a stand-alone initiative.

▶ Watch on YouTube

My Reflection

Before exploring this topic, I tended to think about AI and diversity mainly in terms of algorithmic bias. I now see that the issue is broader.

Technology can create new forms of discrimination, but it can also remove barriers that existed long before AI. The OECD examples of live captioning, image recognition and accessible workplace tools were particularly important because they show how AI can help redesign work around a wider range of human needs.

At the same time, I learned that a system does not have to intentionally discriminate in order to produce unequal outcomes. Historical data, occupational patterns and apparently neutral design choices can disadvantage particular groups.

The gender difference in exposure to generative AI is a good example. If women are more concentrated in highly exposed occupations, an organisation could introduce the same technology for everyone and still create a greater transition burden for female employees. Equality therefore sometimes requires more than treating everyone identically.

For me, this demonstrates why HR professionals need to be involved in technological decisions. The question is not simply whether an AI system is accurate. HR should ask who benefits, who carries the risk, who has access to training, who may be excluded and whether employees can challenge the system.

The fairest workplace will not be the one that uses the most AI. It will be the one that uses technology deliberately to widen opportunity while protecting human dignity and voice.

💬 Join the Discussion

Do you think AI is more likely to reduce workplace discrimination—or make existing inequality harder to see?

What safeguards should organisations introduce before allowing AI to influence recruitment, promotion, performance or access to development opportunities? Share your view in the comments.

References

Brewster, C., Sparrow, P., Vernon, G. and Houldsworth, E. (2017) International Human Resource Management. 4th edn. London: CIPD.

CIPD (2026) Equality, diversity and inclusion (EDI) in the workplace. 11 February. Available at: https://www.cipd.org/en/knowledge/factsheets/diversity-factsheet/ (Accessed: 12 August 2026).

Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K. and Troszyński, M. (2025) Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140. Geneva: International Labour Organization. Available at: https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure (Accessed: 12 August 2026).

International Labour Organization (2025) Global case studies of social dialogue on AI and algorithmic management. 10 July. Geneva: ILO. Available at: https://www.ilo.org/publications/global-case-studies-social-dialogue-ai-and-algorithmic-management (Accessed: 12 August 2026).

OECD (2025) Artificial Intelligence and the Labour Market in Korea. Paris: OECD Publishing. Available at: https://www.oecd.org/en/publications/artificial-intelligence-and-the-labour-market-in-korea_68ab1a5a-en/ (Accessed: 12 August 2026).

OECD (2026) AI and work. Available at: https://www.oecd.org/en/topics/ai-and-work.html (Accessed: 12 August 2026).

U.S. Equal Employment Opportunity Commission (2024) Employment Discrimination and AI for Workers. Available at: EEOC guidance (Accessed: 12 August 2026).

Varma, A. and Budhwar, P.S. (2014) Managing Human Resources in Asia-Pacific. Abingdon: Routledge.