Messaging matters to eBay chief technology officer Mazen Rawashdeh when he talks about the predicted productivity gains from artificial intelligence. Instead of being enthusiastic about AI, some of his colleagues are apprehensive.
“A lot of people get nervous around AI,” says Rawashdeh. “We shift how we talk about gen AI: This is a tool that is going to help you enjoy the job you do.”
It's the balancing act that executives at many companies are performing these days as they adopt AI. How do you get everyone on board with a technology that many fear will replace them?
When talking to his software development team about eBay’s generative AI tools, for example, Rawashdeh says he focuses less on those efficiency gains and more on their career development opportunities. Gen AI can elevate the programming skills of junior and mid-level engineers more quickly than was possible before.
A few times during our conversation, Rawashdeh predicted that eBay's software developers will be 15% to 20% more productive within the next two years thanks to AI. To achieve those gains, eBay has relied partly on GitHub Copilot, an AI-enabled developer tool that can autocomplete code for programmers so they can do their jobs faster.
But the code the technology produces can be a bit generic, says Rawashdeh. The AI doesn't necessarily understand how eBay does things.
That led Rawashdeh to create eBay Coder, which relies on an open-source large language model that's trained on over 250 million lines of eBay code. This generative AI tool is better at helping with tasks like code migration or transferring software code between different operating systems or frameworks.
Beyond boosting productivity for developers, Rawashdeh says there are three other generative AI priorities he's focusing on: improving customer service, scaling the company’s tech infrastructure, and gleaning better insights from customer data to improve the online shopping experience for them.
With nearly three decades worth of customer data in the vault, eBay has a lot of data to sort through. At any given time, there are more than 2 billion product listings on the company’s site and more than 130 million active buyers and sellers. EBay has been using AI to improve search, create more personalized ads, and much more recently, make it easier for sellers to list items on the platform.
Among those newer tools is its "magical listing" feature, which was introduced last year and lets sellers upload a photo and then rely on AI to generate the description to go along with the product listing. Generative AI also recommends listing prices and shipping costs, though the seller has the final say on everything that appears in the posting's final form. And in June, eBay unveiled a new feature that lets sellers replace image backgrounds with an AI-generated backdrop. An image of soccer cleats sitting on a drab living room floor, for example, can be altered to appear in a more visually appealing grass field.
Rawashdeh says that before generative AI, eBay had billions of signals about what consumers bought and sold, and the feedback that they’d share with eBay about their experience on the site. But it was impossible for his teams to look at that data in its entirety and make good decisions, he said. Generative AI, he believes, will increasingly help eBay sort through that massive data trove and find useful insights.
EBay is agnostic about who it will work with for generative AI. It partners with Microsoft for GitHub and buys AI data center infrastructure from Nvidia, for example. Rawashdeh says he decides based on what will create the greatest value for customers. From there, he sorts out whether to buy or build software and, if necessary, then decide what vendor to work with.
"My bias is always towards build," says Rawashdeh, who says eBay relies on a mix of commercial and public large language models and builds its own LLMs.
Rawashdeh is a “boomerang” employee at eBay. He first joined the retailer in 2003 but left in 2011 to work for Twitter, now known as X. He was lured from eBay by the social media company’s fast growth and assorted challenges related to tech and workplace culture that he wanted to help untangle.
After nearly five years at Twitter, Rawashdeh says he left to put more effort into investing in startups and serve as a board member. His intention was to retire, but he quickly returned to eBay in 2016 for his second stint. That also led him to fulfill a dream he had when he first joined eBay.
“I wanted to be a CTO,” says Rawashdeh, who finally became one at eBay in 2019. “That was one of my goals.”
John Kell
Send thoughts or suggestions to CIO Intelligence here.
