Candi Castleberry: Embracing the Future

Amazon’s Candi Castleberry is on a mission to leverage machine learning to build inclusive products and experiences at Amazon—and beyond

by Jackie Krentzman

Imagine you are charged with packaging an Echo Frames device to ship to customers worldwide. How do you ensure that those who might be visually impaired have an accessible experience from the moment they open the package? This is a small example of the very large problem that Amazon’s Candi Castleberry is working to solve.

Castleberry, the company’s vice president of Inclusive eXperiences and Technology (formerly Global Diversity, Equity, and Inclusion), is responsible for a fair and responsible experience in each transaction made by Amazon’s hundreds of millions of customers. Her team works hand in hand with businesses throughout the organization to shape how the world’s fourth-largest company interacts with its breathtaking diversity of customers in 62 countries and, as it does so, maximizes inclusiveness.

At many, if not most, companies, the job of figuring out how to interact with customers and employees is done by engineers, who use rapidly developing machine-learning technology to make and sell experiences and products. However, this task largely happens without taking into account the specific needs and cultural mores of different customer and employee demographics. That can lead, unintentionally, to alienated customers and employee turnover.

Castleberry is on a mission to leverage AI to embed inclusion tech within Amazon but also, in the process, is building a road map that organizations across the globe can harness to build inclusive products and experiences that serve everyone. She believes that if she and her team succeed, diversity, equity, and inclusion practices will grow and thrive everywhere.

“Generative AI is here, and it is the future,” she says. “The world is changing, and we’ve spent decades working with humans, and now is the time for us to see AI not as a threat but as the most important tool for creating the change that is at the heart of our mission.”

The technology’s rapid evolution prompted an urgent look at issues that Castleberry and Amazon addressed by setting in motion a plan to incorporate inclusion processes into its software.

“People knew Gen AI models were being built while we were continuing to partner with people. And while many have concerns about AI, this is the moment where we should shift our focus to embracing these models.”

At Amazon, Castleberry has found an ideal arena for this work. Amazon’s size and scale give it the reach and resources to embed inclusivity in an immense array of products and experiences. “With our scale, our technology, and product delivery, Amazon is uniquely positioned to be the most inclusive company in the world.”

The first step toward realizing that grand ambition was building and implementing an inclusive tech framework internally, with employee-facing products, before taking it to customer-facing products.

But size and reach alone are not the determining factors. Perhaps even more important, Amazon is deeply dedicated to the work. Its commitment comes from the top and is baked into every corner of the company. To underscore its commitment, Amazon is continuing to invest in Inclusive eXperiences and Technology (IXT). The team is tasked with accelerating Amazon’s vision to create inclusive, accessible, and equitable experiences enabled through technology.

“I came to Amazon,” Castleberry says, “because I wanted to work with leaders who want to get it right, and who are committed to staying the course and making the modifications to get it right.”

This thinking is not new for Castleberry. Her work today is the evolution of a theory she developed nearly 20 years ago and has refined multiple times since. She pioneered the concept of moving from “bolted on” and “built in” inclusivity. At Amazon, the work expanded to “born inclusive” products. “Bolted on” refers to programs that sit on top of existing systems to foster equitable outcomes. “Built in” are mechanisms and interventions, such as nudges, that are built into existing products to prevent potentially inequitable outcomes. “Born inclusive” means equity and inclusion are designed into the architecture of a product or system from the start. Thus, Amazon’s IXT team is collaborating with the science and technology team to harness the enormous potential of AI to make sure every product and every experience that Amazon creates is born inclusive.

But that’s not all. Castleberry’s vision extends well beyond Amazon. She sees her company as a lab of sorts, to develop a flexible and nimble model that organizations can adapt—in fact, must adapt—for their organization to avoid being left behind as machine learning becomes the norm.

“It’s been a tough time the last couple of years. The goal is for our work at Amazon to inspire the industry so they may avoid turning to bolted-on solutions.”

What do inclusive tech and born-inclusive products look like on the ground, and how are they generated?

An inclusive technology expands on our learnings from born accessible products and looks towards creating products that are personalized not only to an individual’s browsing history, but also to their culture and identities, by integrating significant nuance into products. It means not producing ads in color combinations offensive to some cultures or ads with overtones at odds with the product or the company’s values. It means delivering targeted products that acknowledge who the recipient is and what is important to them.

It also means creating products that have a social benefit.

“We are uniquely positioned to have the ability to think about how best to deliver inclusive grocery services and health care,” Castleberry says. “Who doesn’t want access to pharmacy delivery with the same mechanisms and capability that you have when ordering food and getting it delivered?”

Such tightly tailored experiences and products can have a huge downstream impact. For example, Amazon’s ability to partner in the past few years with companies putting rockets into lower orbit to provide broadband in places where fiber-optic cable can’t reach will, in time, help someone create a small business selling products long unavailable to rural or isolated communities.

The potential impact is massive. Given the incredible diversity of societies and cultures Amazon serves—let alone all the products it sells—how can a team deliver on the promise of inclusivity?

“We begin by asking, What kind of cultural context do you need in order to build a product that serves a broader audience?” Castleberry says. “Typically, engineers say, we design to spec. They are told, ‘We want this thing, by this time’. No one tells them to avoid such-and-such color because it may be offensive to a significant market segment. No one tells them to avoid certain language that can’t be modified for accessibility. So, we partner with the engineering team to guide them in product development from the time of ideation to best guarantee inclusive technology.”

Castleberry acknowledges that it is impossible to anticipate in real time every need for accessibility and culturally appropriate imagery and language for every community it serves. However, in formulating born-
inclusive products and experiences, a framework emerges that can be expanded upon, within Amazon and beyond.

When Amazon develops a product that can be used by people with visual impairment in, say, both Boston and Bangalore, a universal, baseline born-inclusive standard could be created. Over time, the standard is codified, disseminated, and made public, which means that engineers elsewhere within Amazon or at other companies would no longer need a person sitting beside them as they code, or someone to call the DEI or accessibility team at every milestone check along the way. The product and its delivery system would be continually refined.

“Standards are standards, and they don’t have to be limited to Amazon,” she says. “Our standards can become US standards. They can become global standards. They can be modified by other people who make the standards better. We don’t want inclusive technology to be limited to what you get from Amazon.”

It is that global potential which drives Castleberry. The work, in short, is personal and the culmination of a career that began in traditional DEI practices and has morphed into so much more.

“I feel like I owe it to the [DEI] industry,” she explains. “I owe it to all the people who started this work before me. Because as we think about the shift from a few people with resources influencing the way a product is built, to an iterative born-inclusive technology with an exponential reach and ability to integrate the technology everywhere, the potential of fulfilling our vision of a comprehensive, inclusive business- and human-scaled model is immense.”

Given the scope of the endeavor, when will born-inclusive standards become a reality? Ideally, at the speed of technology. AI is developing and being disseminated so quickly, Castleberry believes, that standards will be in place within two years. “We are on the verge of a huge breakthrough, a sea change in making technology experiences inclusive.”

And she wants all her colleagues to share in this breakthrough. If they can work with their organization’s leadership and engineering teams to apply inclusive standards, these can be preserved and expanded around the world.

She acknowledges it won’t be easy. It will require a significant mind shift by the industry, leaders, and practitioners, from fear of the technology to embracing it and recognizing its inevitability and vast potential for creating more inclusive and equitable workplaces, products, and experiences.

Castleberry is a tireless evangelist, in it for the long haul.

“AI is the future of inclusive experiences,” she says. “It is my mission to make it so.” DW