AI Doesn't Discriminate, Usage Does
Last week, I received an email with one line in bold: “Usage of AI is strictly discouraged as our focus is to understand your thinking and approach.” That line captures a much broader sentiment surrou
By: Raina Roy
Last week, I received an email with one line in bold: “Usage of AI is strictly discouraged as our focus is to understand your thinking and approach.” That line captures a much broader sentiment surrounding artificial intelligence today. Sceptics worry that AI is sexist, biased, and capable of reinforcing existing inequalities (UN News, 2026). Others argue that AI is a shortcut that erodes critical thinking, stifles creativity, diminishes skill development, and threatens intellectual autonomy, an effect researchers have termed “cognitive offloading” (MDPI, 2026). Major companies have acted on these worries. JPMorgan Chase, Apple, and Spotify have all restricted employees’ use of AI outright (CNN Business, 2023).
But why do we limit a tool designed to enhance human productivity? From an economic perspective, AI is an innovation that lets people achieve more with the same time and effort. Economists would say it shifts the indifference curve outward, raising efficiency without raising resource use. In India, where gender gaps run across corporate structures, a tool that is genderless by architecture could improve equity. A large language model does not decide anything based on who is prompting it. A calculator returns 25 for 5 times 5 no matter who presses the keys, and an LLM answers the same way, which makes it one of the more gender-neutral tools available to the workforce.
The equity problem does not arise from AI itself. It arises from who gets to use it. When more men than women engage with AI, the data and language that models train on tilt toward male usage patterns and professional needs (Ho et al., 2025). Over time, the model’s baseline context shifts. This is not an inherent property of the technology. It is a consequence of who gets to use it. Restricting AI does not neutralise this risk but amplifies it because restrictions rarely fall equally.
In the modern economy, AI is a reality. Employers and workers would do better to adapt than to resist, treating AI as leverage rather than as a threat.
India’s Comparative Advantage in AI
India leads the world in AI skill penetration, and the gender data makes the equity opportunity compelling. The government has launched the IndiaAI Mission and established Centres of Excellence for AI, strengthening the country’s research and innovation base. Policymakers have tied these programmes to the vision of Viksit Bharat by 2047, under which India aspires to become a global AI powerhouse.
India ranks 1st globally in AI skill penetration with a score of 2.8, ahead of the US (2.2) and Germany (1.9). India’s AI talent concentration has grown by 263% since 2016. (Stanford AI Index, 2024)
India also leads in AI skill penetration for women, scoring 1.7 against the US at 1.2 and Israel at 0.9. This means that when Indian women access AI, they adopt it at a rate that outpaces that of women in the world’s most advanced economies. (Stanford AI Index, 2024)
Yet a strange contradiction sits alongside these wins. We celebrate AI adoption nationally and discourage it organisationally. Companies announce AI strategies, then email employees telling them not to use ChatGPT. In universities, professors worry about AI-assisted assignments even as their departments launch AI courses. Recruiters say they want tech-savvy employees, yet run AI-averse hiring processes. Investing in highways while banning cars would look much the same. A century ago, literacy meant reading and writing. Today it increasingly means the ability to collaborate with intelligent systems. Institutional resistance to AI is not neutral. It exacts its highest cost on those already structurally disadvantaged.
The Cost of Pessimism
Women face a real and well-documented risk in the global labour market. A 2025 ILO analysis with Poland’s National Research Institute found that in high-income countries, jobs at the highest risk of generative-AI automation account for 9.6% of women’s employment against 3.5% of men’s, nearly three times the exposure. Clerical and administrative roles drive that gap, because women disproportionately hold them and because they rank as the most exposed occupational category. Unless employers respond deliberately, through retraining, redeployment, and policies that route women into AI-augmented rather than AI-replaced roles, AI could meaningfully erode female labour-force participation. The risk raises a design question: how should we build institutions and workplaces so that AI becomes a net gain for women rather than a source of displacement?
A GLO study (2024) shows what happens when institutions leave that question unanswered. Where AI tools are available but not officially sanctioned, women use them significantly less than men, at 45.1% against 57.7%, and associate the usage with misconduct or cheating. A CNBC survey (2026) finds the same instinct at work: 41% of women feel that using AI at work is cheating, compared with 34% of men. Women also tend to seek higher competence before embracing a new technology, while men explore AI at lower proficiency levels. Heavy institutional investment in AI has not dislodged this pessimism at the point of adoption, and the divide deepens along lines of risk tolerance, conformity, and confidence. Women, in particular, disengage rather than experiment: Stanford Social Innovation Review’s meta-analysis covered 143,008 people across 25 countries and found that women had 22% lower odds of using generative AI than men. Scepticism about AI’s value did not drive that gap. Lower confidence did, alongside a perceived need for more training before the tool could pay off.
From “Cheating” to “Fluency”
The GLO research also identifies a turning point. When the employer formally provides the AI tool, usage rates equalise. Official endorsement of AI use reclassifies it for women, transforming it from a suspected form of cheating into a skill-enhancing tool worth capitalising on. Companies that make space for AI in hiring standards and workflows would stop penalising caution, which often rests on the assumption that AI use is unethical, and start rewarding the people who use AI to sharpen their skills. AI governance could then advance gender equality rather than merely avoid harm.
The same study found that women who used an employer-sanctioned AI tool reported measurably higher confidence. The authors link that confidence to a greater willingness to enter competitive, merit-assessed settings. Hiring and promotion decisions turn on exactly that kind of high-stakes evaluation. Niederle and Vesterlund established the baseline problem years ago: equally skilled women opt into competition less often than men do.
AI also appears to lift women’s performance more than men’s. The law of diminishing marginal productivity explains why. AI adds far more marginal value to a non-technical worker than to a seasoned technical veteran. A male software engineer might gain an incremental bump in coding speed. A woman historically shut out of legacy technical syntax can use an LLM to bypass years of structural barriers, automating processes, building tools, and commanding digital infrastructure in natural language. The skill premium on AI use, in other words, runs structurally higher for women. Equal access here produces unequal gains, and the gains favour those previously excluded.
Leading by Example
Some companies have already integrated AI rather than fenced it off. Unilever cut 70,000 person-hours of manual interviewing and assessment, made its job descriptions more inclusive, and attracted a balanced talent pool by combining AI with a blind application evaluation process. Women now hold 51% of its management roles globally (NITI Aayog, 2025). Young et al. (2023) argue that generative AI offers a rare chance to disrupt traditionally male-dominated fields and to build diversity into them early.
Way Forward
Firms that never officially encourage AI widen the gender gap, which makes institutional endorsement a structural imperative rather than a soft preference. When the firm provides AI as a baseline tool, women engage competitively at the same rate as men. Endorsement dismantles barriers that have long gatekept important roles.
The private sector, and the technology industry in particular, carries an outsized role in making AI adoption equitable (UNCTAD, 2025; UN Women, 2025). UN Women (2025) and UNESCO’s Recommendation on the Ethics of AI set the regulatory floor, calling on private organisations to adopt responsible practices: to strip gender bias from models, to increase women’s representation in AI leadership, and to build safeguards against AI-amplified violence.
Which brings me back to that email. “Usage of AI is strictly discouraged” was meant to protect something: original thinking, presumably. But it will not. The people who ignore it will be those already confident enough to experiment, and the GLO and Stanford numbers show who that is. A ban does not measure how well anyone thinks. It measures who was willing to risk being caught, and it sorts a workforce accordingly.
About the author: An economics and policy enthusiast with an MSc in Development Economics and Policy, exploring the intersections of gender and labour economics. Passionate about research with experience in NGO work and student-led policy initiatives. Beyond economics, she has spent over 20 years practicing classical dance.