Why the next workplace inequality may not come from biased algorithms—but from how organisations implement AI.

For the past two years, organisations have focused on an important question:

Can artificial intelligence discriminate?

The evidence suggests it can.

Research has shown that AI systems can reproduce or amplify bias embedded in training data, system design and human decision-making. Regulators around the world, including the Australian Human Rights Commission, have rightly highlighted the need for fairness, transparency, accountability and meaningful human oversight when AI influences decisions about people’s lives.

But another question deserves equal attention.

Can the way organisations implement AI create new forms of workplace inequality, even when the technology itself is unbiased?

I believe the answer is yes.

The next workplace divide may not be created by algorithms alone. It may be created by the vastly different opportunities employees receive to learn, use and benefit from AI.
If some employees have access to better tools, better managers, better training, protected time to experiment and opportunities to apply AI to meaningful work, while others receive little more than access to the software, organisations risk rewarding unequal opportunity as though it were unequal talent.

That is not primarily a technology issue. It is an organisational design problem.

Unless leaders address it deliberately, the AI era could unintentionally create a new form of workplace inequality—one built not on who people are, but on who organisations choose to enable.

We already understand algorithmic bias

Much of the public debate about AI and fairness has focused on bias embedded within AI systems. That concern is entirely justified.

Research has consistently shown that AI can reproduce or amplify historical patterns of discrimination when training data, system design or implementation reflect existing human biases. The Australian Human Rights Commission has warned that these risks may affect employment, financial services, healthcare and many other high-impact decisions. Their work has reinforced the importance of fairness, accountability, transparency and meaningful human oversight throughout the AI lifecycle.

These risks require careful governance. But they are only part of the picture. An organisation could deploy an AI system that is technically robust, carefully tested and subject to appropriate human oversight—and still create unfair outcomes.

Not because the technology discriminates. But because the organisation creates unequal opportunities for people to develop the capability to use it effectively. That possibility deserves far greater attention than it currently receives.

The real risk is unequal enablement

As AI becomes embedded in everyday work, organisations are not simply introducing new technology. They are redistributing opportunity.

Some employees receive access to high-quality AI tools, practical training, supportive managers, protected time to experiment and psychologically safe environments in which mistakes become learning opportunities. They are encouraged to test ideas, refine workflows and combine AI capability with professional expertise and sound judgement.
Others receive access to the same software—but little else. They may be uncertain about what is permitted. Concerned about making mistakes. Excluded from pilot programs. Given little opportunity to practise. Or expected to develop AI capability in their own time.

On paper, everyone has access. In reality, not everyone has the same opportunity. That distinction matters.

Capability develops through opportunity—not simply access to AI systems.

The OECD has observed that the benefits of AI-related training will not be distributed evenly unless organisations address inequalities in learning opportunities, infrastructure, data and relevant skills. Likewise, the International Labour Organization and the United Nations have warned of an emerging global “AI divide”, driven not only by unequal access to technology but also by unequal opportunities to develop AI capability.

This is where a new form of inequality begins to emerge. Not unequal access to AI.

But rather unequal opportunity to exercise the full benefits of AI

The new AI divide

The original digital divide was largely about access to technology.

The emerging AI divide is far more complex.

It is shaped by differences in:

• access to appropriate AI tools
• the quality and timing of learning opportunities
• managerial encouragement and role modelling
• protected time to experiment safely
• clarity about appropriate and inappropriate uses
• opportunities to influence how AI changes work
• access to increasingly sophisticated AI-enabled assignments
• the ability to combine AI with professional expertise and human judgement.

Recent OECD research reinforces that workers’ ability to benefit from AI depends not simply on whether technology is available, but on whether they possess the skills required as jobs and tasks evolve.

This distinction is critical. Employees who receive stronger support often complete work more efficiently, contribute to more complex projects and become recognised as early adopters. That recognition rarely ends there.

It often leads to more visible assignments, greater exposure to senior leaders, broader networks, increased confidence and accelerated career development.

Meanwhile, employees who have not received the same opportunities may appear less adaptable, less innovative or less productive—even when the underlying difference is not capability or commitment, but unequal access to learning and support.

The danger is that organisations begin rewarding the outcomes of unequal enablement as though they were evidence of unequal potential.

AI adoption is a leadership challenge, not a technology project

Too often AI implementation is treated as a technology initiative. In reality, it is a leadership, workforce and organisational design challenge.

Technology changes quickly. Capability develops more slowly. Organisations therefore need to think beyond software deployment and focus on how work itself is changing.
OECD workplace research suggests that the impact of AI depends heavily on how implementation is managed—including how jobs are redesigned, how new skills are developed and whether employees are actively involved in the transition.

If AI capability increasingly influences performance ratings, promotion decisions or access to high-value work, then existing differences in opportunity may become amplified.

For example:

• employees in some functions or locations may receive earlier access to AI tools
• flexible or part-time employees may miss informal learning between colleagues
• employees with significant caring responsibilities may have less capacity to undertake optional learning outside work
• some managers may actively encourage experimentation while others discourage it
• employees with greater confidence or previous experience may receive the most sophisticated AI-enabled assignments
• organisations may reward individuals who independently develop AI capability without considering who had the time, opportunity or support to do so.

None of these outcomes are inevitable consequences of AI.They are consequences of organisational decisions and that means they are also within an organisation’s control.

Fairness requires more than equal access

Many organisations will understandably respond:

“Everyone has access to the same AI tools.”

But equal access is not the same as equal opportunity.

Providing every employee with the same software does not guarantee that everyone can use it with the same confidence, competence or effectiveness.

Fairness means designing systems, processes and opportunities so that employees have a genuine opportunity to develop, contribute and succeed. That does not necessarily require identical interventions.

Different roles and different groups may face different barriers- some employees may require foundational AI education; others may need role-specific application; some may benefit from coaching. Some may require more accessible learning formats, clearer governance or greater opportunities to practise within their everyday work.

Managers themselves may need support to redesign work, assess AI-assisted performance and create psychologically safe environments where experimentation is encouraged rather than punished.

What is needed is Universal goals. Targeted strategies.

The goal remains the same for everyone.

The support required to achieve it may differ.

The objective is not special treatment.

It is removing barriers that prevent people from contributing at their full potential.