AI Won't Eliminate Workplace Bias, Unless We Design It To

Written by: Sophie Warwick

Like many people, I've found myself thinking a lot about AI lately. How will it change the workplace? Which jobs will it replace? And, if we're being honest, should I be at least a little concerned about the robots eventually taking over? Jokes aside, I think these are important questions.

There's an optimistic part of me that remembers we've been here before. During the Industrial Revolution, many feared machinery would eliminate jobs altogether. While there was significant disruption, technology ultimately changed work more than it eliminated it. Some roles disappeared, but entirely new industries and careers emerged. We've continued to see this throughout history. Technology automates certain tasks, allowing us to spend more time on work that creates new value and focus on work that simply wasn't possible before.

I've seen that evolution even within my own lifetime. I was 14 when I got my first cell phone. I had about 50 text messages a month, and my heart would briefly stop every time I accidentally hit the internet button because those early data charges were astronomical. Today, my phone's primary function is arguably not even being a phone. I use it for email, maps, banking, weather, taking photos, and looking up random facts far more often than I use it to make a call.

Technology evolves, and we adapt. What feels different this time isn't that technology is changing, it's the speed. I left on parental leave at the end of 2024 and returned seven months later. I expected to spend a few weeks remembering where everything was saved and getting back into the rhythm of work. What I didn't expect was that the way people worked had fundamentally changed.

When I left, AI and large language models already existed. They weren't new technologies, but they still felt like tools people experimented with occasionally. By the time I returned, AI had become woven into everyday work. It was helping people draft emails, summarize meetings, write code, brainstorm ideas, prepare presentations, and analyze information. In what felt like the blink of an eye, AI had become as commonplace in many workplaces as email.

That transformation is exciting. But it also raises an important question. When we talk about the integrity of AI, we spend a lot of time talking about factual accuracy. Is the information correct? Has it hallucinated? But what if the information is factually correct and still ethically wrong?

Imagine an organization uses AI to help identify employees who are ready for promotion into leadership roles. The system is trained on years of historical promotion decisions. If, historically, fewer women were promoted into leadership positions, the AI may begin to identify fewer women as "leadership ready" simply because that's what the historical data suggests leadership has looked like. The recommendation may even appear statistically sound. But it's also reinforcing the very inequity the organization is trying to address.

The same challenge exists in performance management. If historical performance reviews consistently describe men as "confident," "strategic," or having "executive presence," while women demonstrating similar behaviours are more likely to be described as "collaborative," "supportive," or "hardworking," an AI system trained on those reviews can begin to associate leadership potential with traditionally masculine language. Again, the AI isn't creating bias. It's learning it.

This is where I think the conversation around AI needs to evolve. The question isn't simply whether AI is correct. It's whether it's fair. We're already beginning to see evidence of this challenge. A study by researchers at Hong Kong Polytechnic University and Peking University asked more than 1,000 engineers to evaluate identical pieces of code (There’s a Reason Women Aren’t Swooning Over AI Like Men Are, 2025). The only differences were the perceived gender of the author and whether AI assistance had been used. Engineers who used AI were generally rated as less competent. But the competence penalty wasn't equal. Women experienced roughly twice the reduction in perceived competence compared with men for identical work. That finding really struck me.

Women are often criticized for not keeping up with new technology or for being slower to adopt. Yet when they do embrace it, they're judged more harshly for using it. It's another example of the double standards that already exist in our workplaces finding their way into emerging technologies.

This is why I believe AI has enormous potential, but only if we're intentional about how we design and use it. It learns from historical hiring decisions, compensation data, promotion patterns, performance reviews, and countless other datasets that reflect the world as it exists today, not necessarily the world we want to build tomorrow. If we aren't careful, AI won't eliminate workplace inequities. It will automate them.

None of this means we should stop using AI. Far from it. I use AI regularly, and I'm continually impressed by how much more efficiently it allows me to work. It frees up time for deeper thinking and the parts of my work that I enjoy most. Used thoughtfully, I believe it has the potential to make us significantly more productive. But we shouldn't confuse efficiency with equity.

As organizations continue integrating AI into hiring, compensation, performance management, succession planning, and promotion decisions, we have a responsibility to ask harder questions.

• Whose data trained this model?

• What biases might it be reinforcing?

• Who benefits from its recommendations?

• And who could be unintentionally disadvantaged?

The future of AI won't be determined by the technology alone. It will be determined by the choices we make as humans. If we're thoughtful, AI has the potential to help build more equitable workplaces. It can reduce administrative burden, increase access to information, and allow us to focus more of our time on meaningful, human work. But if we're careless, it risks accelerating the very inequities we've spent decades trying to eliminate.

Technology is only as equitable as the systems it learns from and the people who choose how it's used. Integrity isn't just about whether AI gives us the right answer, it's about whether it's helping us build a fairer future. Correct and fair are not always the same thing.