How eBay is using generative AI, computer vision to enhance CX

Ecommerce giant eBay aims to revolutionize its marketplace by integrating generative AI and computer vision. With these technologies eBay aims to enhance its understanding of customer preferences and deliver highly personalized shopping experiences.

During the VentureBeat Transform 2023 conference, Nitzan Mekel-Bobrov, chief AI officer, and Xiaodi Zhang, chief product officer, delved into eBay’s ambitious plans to scale its already robust AI infrastructure. With the power of generative AI and decades of accumulated data — comprising billions of images, customer interactions and item details — the company aims to stay at the forefront of technological innovation.

>>Follow all our VentureBeat Transform 2023 coverage<<

“We’ve doubled down on our investment in generative AI and computer vision technologies because we believe that [they have] transformational value for impacting the customer experience,” Mekel-Bobrov told VentureBeat. “We’ve also been working on building internal generative AI tools to aid the productivity of our developers, analysts and data scientists.“ 

Using generative AI to enhance buyer-seller experience

The company said it is pursuing innovative approaches to enhance the buyer experience through AI-driven product discovery. 

By using generative AI throughout its platform, eBay aims to address challenges and streamline the purchasing process for both buyers and sellers. It wants to create captivating ways purchasers can discover products both during the purchase moment and at earlier stages of the buying journey, boosting user satisfaction and eliminating obstacles users encounter.

Zhang emphasized the company’s significant investments in improving the selling experience and listing flow. Zhang explained, for example, the challenges that arise during the listing process, such as determining item descriptions, pricing and the appropriate level of information. 

Recognizing this as a valuable opportunity, eBay aims to use generative AI to address the “cold start” problem, for example. This refers to how newly added items lack interactions that can draw attention to them. Improving listing flow should facilitate faster and more effective selling for individual consumers as well as business sellers.

The need for AI governance

eBay emphasized data privacy and ethical development as fundamental principles of integrating generative AI into business functions.

According to Mekel-Bobrov establishing governance and implementing clear guardrails enables individuals to navigate among safe and permissible areas within the organization.

To address these concerns, the company has recently established an Office of Responsible AI, which includes cross-functional representation and influential thought leaders in the field. The team’s primary focus is tackling issues such as bias measurement and hallucination assessment, specifically in the context of responsible generative AI.

Mekel-Bobrov elaborated on the challenges associated with the underlying data corpus used in historic modeling campaigns. These models often exhibit biases regarding racial representation, diversity and visual aspects such as body type and image. 

“We are actively exploring how to leverage these foundational models while accounting for their limitations, including the lack of diversity,” he said. “For us, it’s not just about solving the immediate problems that are very obvious, but it’s also about looking at the harder problems that will take some research to figure out solutions for.”