Can AI thrive without neurodivergent talent?

Earlier this year, Alex Karp, CEO of software company Palantir, said there are two paths to a successful future in the age of AI: vocational training or being neurodivergent. As co-president and CTO of Understood.org, a leading nonprofit that supports millions of neurodivergent people in the US, I see why this sentiment resonates. But it doesn’t tell the whole story.

It makes sense to position divergent thinking as an advantage amid real concerns of AI being the end of originality. When everything starts to look or sound the same, people who see and experience the world differently bring outsize value. 

Research is still in its early stages. But studies suggest that neurodivergent people may be more likely than neurotypical people to use and benefit from AI tools, like chatbots.

Neurodivergent minds bring strengths that AI development urgently needs like pattern recognition, nonlinear thinking, and a tendency to question assumptions. When those perspectives are missing, gaps become embedded into products, programs, and outcomes. AI reflects both the data it’s trained on and the people who shape it so that’s why cognitive diversity at the point of creation is critical. 

AI’s real power users are neurodivergent

Understood.org’s latest survey of over 2,000 US adults finds that 78% of neurodivergent employees report using AI tools at work, versus 59% of neurotypical employees. This echoes reports from EY that neurodivergent professionals are significantly more likely than neurotypical professionals to use AI daily.

But the significance isn’t just higher usage, it’s how that usage looks in practice. When someone regularly uses a tool, they really interrogate it. They see where it helps, where it falls short or introduces friction, and where the system still assumes a default user who doesn’t exist.

That’s what makes neurodivergent users the closest thing AI has to true power users today. They can push tools into edge cases and encounter breakdowns others never reach. And they find surface gaps between how systems are designed, and how people actually use them. The result is a feedback loop the industry can’t afford to ignore. That’s why embedding neurodivergent perspectives from the start leads to better systems.

Teams that think differently are equipped to anticipate a wider range of user needs, reduce biases, and design tools that work in diverse, real-world contexts. When you design for the margins — or even better, when the margins are part of the design team — solutions tend to be stronger for everyone. 

AI doesn’t know what it doesn’t know

Including neurodivergent talent isn’t just about improving AI’s performance. It directly shapes how these systems understand and represent people in the first place. Last September, Amy Gaeta, an AI ethicist researcher, explored how we can build better AI, and the role neurodivergent and disabled communities play in that process.

Ask an AI model to generate or describe a disabled person, and the imagery is strikingly consistent: a person in a wheelchair. Not because that reflects reality, since disabled could refer to a wide spectrum of both visible and non-visible differences. But AI models default to what they’re trained on. If the datasets lack diversity or aren’t representative, and the people curating them don’t recognize those gaps, then the AI learns a narrow definition of disability. And it repeats it at scale.

This can have severe consequences, perpetuating harmful stereotypes, underrepresentation, and misrepresentation in ways that compound over time. What AI is trained to recognize is ultimately what it decides exists.

Build cognitively diverse teams by design

If AI is going to shape how we learn, work, and communicate, then the teams building it must reflect a fuller range of how people think. There’s no shortage of talent to make it happen since nearly one in three US adults identify as neurodivergent. 

But the shift required isn’t just who’s on the team, it’s how their perspectives shape what gets built and what decisions get made. Start treating cognitive diversity as a design input across the teams building these systems. Develop better training data that reflects these perspectives from the start. Creating environments where neurodivergent minds can thrive should be the norm.

The future of AI is neurodivergent

We’ve reached an inflection point. Lack of awareness of neurodiversity is no longer the barrier. Action is. Neurodivergent perspectives have to shape the foundation, not be layered on after the fact.

If we don’t train AI on the breadth of human experiences, we’re building worse, less capable systems as well as misrepresenting communities. So this isn’t about representation for its own sake. Technology will only work as well as the range of experiences it reflects.