Artificial Intelligence (AI) is transforming how we work, hire, and even how we define success—but it’s also quietly reshaping conversations around diversity, equity, and inclusion (DEI). While AI promises efficiency and data-driven insights, it also poses serious risks when bias goes unchecked. On a recent episode of DEI After 5, I sat down with Dr. Alexandra Zelin to unpack what this means for today’s workplaces—and for the future of inclusive leadership.

The Promise and Peril of AI in the Workplace

AI’s rise has brought undeniable innovation. From streamlining hiring processes to identifying performance trends, organizations are using AI tools to make quicker, more “objective” decisions. But as Dr. Zelin pointed out, objectivity is an illusion if the data behind these systems isn’t diverse or equitable.

AI learns from the data it’s fed. When that data reflects historical inequities—like the underrepresentation of women and people of color in leadership roles—it doesn’t correct the problem; it reinforces it. We’ve seen this play out in hiring algorithms that favor men’s resumes or in medical research where AI models fail to recognize symptoms in women or nonwhite patients because the training data lacked diversity.

Simply put: if the inputs are biased, the outputs will be too.

Why Diverse Data Matters

Diverse data isn’t just a technical issue—it’s an ethical one. When data reflects only a narrow slice of the population, it limits opportunity for everyone else. Dr. Zelin used Amazon’s hiring experiment as a cautionary tale: when the company trained an algorithm on resumes from existing employees (mostly white men), the system learned to favor similar candidates. Instead of broadening opportunity, it replicated exclusion.

This is why diversity in AI data sets is critical. It’s not enough for technology to be innovative—it must also be inclusive. That means bringing in voices from underrepresented groups not just as subjects of the data, but as creators, testers, and decision-makers in the design process.

The Role of History in Modern Data

Data doesn’t exist in a vacuum. Historical context shapes it—and ignoring that context can lead to devastating blind spots. Consider how redlining continues to influence school funding and neighborhood investment, or how standardized tests like the SAT privilege certain cultural experiences. These systemic biases become baked into the data that AI learns from, creating a self-reinforcing cycle.

If we don’t account for those historical inequities, AI will simply digitize discrimination under the guise of neutrality. That’s why inclusive design and critical data review are so important—because fairness isn’t automatic. It has to be built.

Laws Are Catching Up

Some progress is being made. New York City, for example, has passed legislation requiring companies to disclose when they use AI in hiring and to conduct equity audits of their systems. These laws are a step toward greater transparency and accountability, helping ensure that technology doesn’t operate unchecked behind closed doors.

While these regulations don’t yet capture the full complexity of intersectional discrimination, they open the door to necessary scrutiny. They challenge organizations to look beyond surface-level diversity numbers and confront systemic barriers that limit access and opportunity.

AI and Workplace Equity Analysis

Beyond hiring, AI can also be used for good—to uncover inequities within organizations. When trained responsibly, AI can analyze patterns in promotions, pay, and engagement to reveal where disparities exist. It can help organizations ask better questions: Who gets access to stretch assignments? Whose feedback is taken seriously? Who’s being left behind?

But again, AI is a tool, not a cure. It requires human oversight, context, and ethical interpretation. Numbers alone can’t tell the full story of someone’s experience at work. Humans must interpret what the data means—and decide what to do about it.

Human Oversight Is Non-Negotiable

One of the most important takeaways from my conversation with Dr. Zelin is that AI needs human interpretation. Technology can process information at lightning speed, but it can’t understand nuance, empathy, or lived experience. Both humans and AI are capable of bias—the difference is that humans can reflect, adjust, and make meaning.

That’s why the future of inclusive workplaces isn’t about replacing human judgment with algorithms—it’s about using AI to support it. AI can flag patterns and inconsistencies, but humans must provide the context and compassion to respond appropriately.

Building an Inclusive AI Future

AI can either amplify inequality or accelerate inclusion—it depends on how we build and use it. The key lies in:

* Diversifying data sources to ensure AI reflects a wide range of experiences and identities.

* Embedding transparency through regular audits, equity impact assessments, and open reporting.

* Keeping humans in the loop, especially those who understand the social and cultural dimensions of bias.

* Acknowledging history and the systems that shaped today’s inequities.

If we get this right, AI can be a powerful partner in advancing equity and belonging at work. But that starts with leadership that values inclusion as much as innovation.

The goal isn’t just smarter technology—it’s fairer outcomes.