The Real Question

A professional who expects her work to be scrutinized more closely may have good reason to verify an AI-generated claim before putting her name on it. A manager who carries an outsized share of relationship work may notice reputational risks that a productivity dashboard misses. A founder with less access to forgiving capital cannot treat a costly automation mistake as a learning experience.

The better question is not why women lack confidence. It is whether workplaces give them a credible way to turn AI use into visible evidence of judgment, learning, and leadership.

Unlike a list of courses or a badge announcing "AI literacy," it shows what someone can actually do.

The Exposure Gap Is Not Neutral

The ILO's 84-country analysis found that occupations dominated by women are almost twice as likely to be exposed to generative AI as occupations dominated by men. Exposure does not mean inevitable job loss — in many cases, tasks will change rather than disappear. But the distribution matters, because it shapes who is expected to adapt quickly and who carries the risk when organizations introduce tools badly.

Women are also underrepresented in many of the technical roles that design, buy, and govern AI systems. That can produce a lopsided workplace experience: heavily exposed to automated change in administrative, customer-facing, communications, and professional work, while having comparatively little influence over how those systems are selected and evaluated.

Telling women to "lean in" to that arrangement does not solve it. Organizations need to make AI competence legible, without rewarding reckless experimentation.

What Belongs in an Evidence Portfolio

A useful portfolio does not require confidential data or polished case studies. One page per task is enough. Each entry should contain six elements.

1. Name the real task. "Used AI" is not evidence. "Prepared a first draft of a client briefing" or "compared three supplier proposals" is specific enough to evaluate.
2. Record the approved information used. This shows the employee understood data boundaries and didn't feed private or proprietary material into an unapproved system.
3. Preserve the relevant AI output. Not to celebrate the machine's answer — to make the starting point visible.
4. Document the human judgment applied. What was verified, rejected, reframed, or added? This is often the most valuable part, because it reveals expertise rather than mere tool access.
5. Capture the outcome. Did the work become faster, clearer, more accurate? Honest outcomes build more credibility than universal success stories.
6. Note the risk or ethical issue considered. Bias, privacy, transparency, accessibility, IP, or the effect on someone else's opportunity. Leaders who catch these issues early aren't slowing innovation — they're making it sustainable.

Make Judgment Visible

A portfolio changes the meaning of AI skill. Instead of asking who produces the most prompts, it asks who can define a worthwhile problem, recognize weak evidence, and take responsibility for the result.

That distinction matters for women whose expertise is frequently treated as supporting work rather than leadership. Consider the employee who catches a misleading claim in an AI-generated sales proposal, notices that an automated shortlist excludes qualified career returners, or redesigns a customer-service workflow so vulnerable users can still reach a person. Those interventions may prevent real damage, yet they often disappear from performance reviews.

It can also make learning safer. Employees don't need to pretend every experiment worked. A strong entry may explain why a tool was rejected for a particular task — knowing when not to automate is itself a form of AI competence.

Use Portfolios in Promotion and Investment Decisions

Employers should incorporate evidence portfolios into development conversations, promotion reviews, and leadership programmes. The standard should not be volume of use — it should be quality of application.

A marketing leader may demonstrate how she used AI to test audience assumptions while protecting customer data. An operations manager may show that a proposed automation saved five minutes per transaction but created seven minutes of exception handling, prompting the team to redesign the process. A lawyer may document how she verified citations and narrowed a tool's role to preserve professional accountability. These examples reveal transferable leadership: problem definition, risk sensing, evidence evaluation, workflow design, and learning from correction.

Founders and investors can use the same approach. For a founder, the portfolio can demonstrate disciplined execution to customers and investors — where automation improved margins, where human review remained essential, and how the company responded to mistakes. That is far more persuasive than a general claim that the business is "AI-first."

Do Not Turn the Portfolio Into Surveillance

Any useful practice can become a burden if implemented carelessly. An evidence portfolio should not become a daily log of every prompt, or a ranking system that rewards constant tool use. It should not expose private employee experimentation or become a pretext for monitoring work.

Participation should focus on selected, meaningful tasks. Employees need clear guidance about acceptable tools and information. Managers should review portfolios for learning and decision quality, not stylistic conformity — and people must be free to record that AI was inappropriate, or that the cost of verification outweighed the benefit. Organizations should also provide time for the work itself; asking employees to document AI use after hours simply adds another invisible task, often to people already carrying disproportionate administrative and relational labor.

Replace Confidence Theatre With Proof

Recent adoption research suggests that gendered perceptions influence how people approach generative AI. That finding should not be converted into another programme designed to fix women's attitudes while leaving workplace incentives untouched.

Confidence grows from credible experience. It grows when people can try a tool on a bounded task, receive useful feedback, correct mistakes without humiliation, and show what they learned. It grows when caution is recognized as judgment, and when successful use creates opportunity.

The future of work does not need women to imitate the loudest early adopters. It needs organizations to recognize the people who can make powerful systems useful, accountable, and humane.