According to the World Economic Forum, only twenty-two percent of AI professionals are women. FutureWeek spoke to Wendy Gonzalez, CEO of Sama, a fifty percent female-led, mission-focused firm that provides responsible data for AI models.
Gonzalez believes that as we witness this pivotal moment in AI innovation, it’s more important than ever for AI companies to prioritise diversity. She shares why clear frameworks, evolving standards, and inclusive representation are foundational to building AI systems that truly serve users.
Why has Sama committed to diversity in the way it has?
Sama is an AI company with a social mission. That social mission is to provide people digital up-skilling from underserved communities and full time employment, so that they can not only build skills of the future, but can sustainably move themselves out of poverty.
We hire at least fifty percent women. We are founded on the belief that talent is distributed equally but opportunity is not. I believe that ties into my own beliefs around diversity and why it’s important – I’m a child of immigrants and grew up in the Seattle, Washington area. Everybody deserves a seat at the table, especially when it comes to global technologies like AI.
Why is it important for AI companies to have an emphasis on diversity?
In general, diversity and bringing people who’ve got different experiences, different skills and different talents to the table inherently creates better results.
It’s incredibly important to have a range of opinions, backgrounds and experiences, so that you can challenge and create better solutions and better products.
Beyond half of our workforce being women, we are also heavily focused in East Africa.
AI doesn’t know a border, or a person, or ethnicity, or a gender, right? These are global, ubiquitous technologies. The challenge of not having some level of diversity or inclusion is that some of these general models don’t understand diverse groups. So, getting more people, and a broader group of people, engaged in the development process is going to make for a better and flexible tool set. The more people you have at the table, the better the product, not just inclusive products, but better products for users overall.
What negative consequences is there for AI companies if there isn’t that diversity?
A lot of these general AI models are built on publicly available information – in its nature this is bias. It becomes really important to have appropriate diversity in the data. There are companies that are leveraging things like RAG embeddings and bringing proprietary or additional data to the table.
If you don’t have enough diversity in the design and development of models, potential risks could be anything from alienating users to ingraining bias in things like financial lending systems, for example. When you don’t have the right people thinking about how to evaluate the performance of a model, bias can get introduced.
What steps do we need to take for there to be more women around these tables?
It starts with founders, investors and leaders to determine what they believe is important. We can’t wait and rely on regulation and policy. It’s ultimately about building a better product and trying to build responsible and trustworthy AI.
Building transparency and standards around AI are all things we can do to encourage those creating AI to keep these considerations in mind. An example of this is responsible AI frameworks.
‘Responsible AI’ isn’t the same as ‘ethical AI’, it’s just a method of development. It starts with the engineers and scientists building these models – if you can provide this guidance to them then you can start to build best practices.
Why do we need trust and transparency to build better models?
Trust and transparency aren’t just compliance checkboxes – they are foundational to developing effective AI systems. Rather than waiting for regulatory mandates, the focus should be on adopting best practices and fostering a culture of education and open communication around what AI is and how it works.
Take, for instance, the common scenario where users sign up for services without reading the terms and conditions. Many people using free versions of AI models are unaware that inputting personal information may contribute to training those models. This lack of awareness highlights a critical need for public education around AI usage and data implications.
Building trust means ensuring users understand what they’re engaging with – and that starts with transparency. Would you step into a self-driving taxi if you didn’t trust its safety mechanisms? Probably not. In the same way, users will only adopt AI technologies when they feel confident in the safeguards, reliability, and ethical standards underpinning them.
How can AI companies ensure their models are ethical?
It’s crucial to establish clear frameworks that developers and companies can follow. These should aim to make the development process more transparent and ethically sound. But beyond technical frameworks, we also need inclusive representation. Diverse voices – including women, who make up half the global user base – must be part of shaping AI. Their inclusion ensures that the technology reflects and serves the full spectrum of human experience. This isn’t just good for representation, it’s good for business too.
In short, trust is earned through transparency, education, and inclusion. These aren’t just ideals – we have to embody them in practice. We must walk the walk.
