Artificial intelligence is often described as a democratising technology. Capabilities that once required specialist teams and significant resources are becoming widely accessible. Knowledge can be accessed faster, complex tasks automated and individuals can now work with capabilities that, only a few years ago, would have belonged to much larger organisations.
But access alone is not inclusion. The more important question is who has the capability and authority to shape how AI is built, deployed and governed. Who determines which problems AI solves? Who decides what constitutes acceptable risk? And who can challenge its output when something goes wrong?
Women remain significantly underrepresented in AI, particularly in leadership. Yet the issue is bigger than representation itself. Representation asks who is present. Inclusion asks who can participate. Agency asks who can decide.
AI is becoming part of the infrastructure of our economies and institutions. It increasingly influences decisions around hiring, lending, risk, capital allocation and the organisation of work. The people designing and governing these systems therefore influence not only the technology, but the economic decisions surrounding it.
Algorithmic monoculture
AI does not emerge independently from society. It learns from our data, language, decisions and historical records. Those records already contain inequalities.
The danger is that underrepresentation becomes self-reinforcing. If women are less visible as leaders, experts and decision-makers, the digital record contains fewer examples of women occupying those positions. AI systems can then influence who is surfaced, recognised, recommended or given opportunities, creating new data from which future systems learn.
For financial institutions and policy-makers, this is not simply a diversity issue. It is also a question of resilience. If institutions rely on similar models, trained on similar data and deployed in environments where decision-makers share similar assumptions, we may create highly efficient systems that are also capable of reproducing the bias at an enormous scale.
We risk creating an algorithmic monoculture, where AI increasingly influences lending, risk assessment, supervision, capital allocation and fraud detection. In this case, diversity of judgement becomes part of institutional resilience.
Addressing this requires action at several levels. At the model level, institutions need appropriate data, testing and continuous monitoring. At the human level, people need sufficient AI literacy and authority to challenge automated recommendations. At the institutional level, boards must understand who builds systems, who validates them, who challenges them and, ultimately, who has the authority to override them.
This becomes even more important with the emergence of agentic AI. As systems become capable of recommending, prioritising and potentially executing actions, human intervention may become less frequent. But human judgement will become more consequential. Inclusion must be designed into an institution’s AI capability strategy, rather than treated as a programme sitting alongside it.
Shifting the dial
The emerging economic divide is no longer simply between those who have access to AI and those who do not. It will be between those who have the capabilities to use AI to create value, exercise judgement and lead transformation and those who do not.
Competitive advantage will therefore depend increasingly on talent density: how many people within an organisation or economy can contribute effectively in an AI-augmented environment. We speak extensively about investing in the infrastructure for AI. We must be equally ambitious about building the infrastructure for talent.
Progress cannot simply be measured by how many women complete AI training. We must ask how many move into AI-related roles, lead transformation projects, participate in strategic decisions and gain authority over the systems shaping their institutions. Women should not simply be users or beneficiaries of AI. They must also be among those building the systems, challenging the outputs, governing the risks and deciding what AI is optimising for.
The question is therefore not whether women will use AI. They already are. It is whether they will have the capability, opportunity and authority to shape what it becomes.
Tram Anh Nguyen is Chairwoman of Global Women in AI and Co-Founder of the Centre for Finance, Technology and Entrepreneurship.
Interested in this topic? Subscribe to OMFIF’s newsletter for more.