Ebay’s Nitzan Mekel-Bobrov has big plans for helping the ecommerce mainstay evolve into what he calls an AI-first company. The December launch of eBay’s proprietary AI-based tech, which can generate 3D product views, is a sign of more immersive shopping and AI-enhanced customer communications to come, built using computer vision, natural language processing, streaming and computer graphics.
As eBay’s chief artificial intelligence officer, Mekel-Bobrov — who joined the company last year after helping lead AI engineering teams at Hearst, Capital One and most recently Booking.com — takes what he calls a distributed approach to disseminating AI across eBay. People in the company’s marketing science, advertising science, search science and buyer experience teams all have domain-specific strategies “but they’re also feeding into the broader enterprise-wide strategy around maturing our AI at eBay and becoming an AI-first company, which is not something any one domain can accomplish on its own,” Mekel-Bobrov told Protocol in an interview this week.
Nitzan Mekel-Bobrov, eBay's chief artificial intelligence officer, makes the rules.eBay
Still, Mekel-Bobrov guards against haphazardly building AI for customer use or incorporating AI-centric tools into workflows without parameters. It’s why he’s creating standards, best practices and governance for the use of low-code and no-code AI tech by others inside the company, and why he says AI requires a unique form of monitoring and maintenance that other software does not.
This interview has been edited and condensed for clarity.
Tell me about what a “distributed model” for AI technology use at eBay looks like, maybe in relation to the new 3D tech.
So that as a capability is something that we’re developing centrally, and then teams across different parts of the customer journey will be able to leverage that capability in an easy-to-deploy way. Our buyer experience team has deployed this on a number of our pages, a number of our portions of the customer journey, that they own. You’ll also see it in our eBay stores.
So far we’ve launched this for sneakers. If you look at some of our top sellers, they now feature 3D in their stores, in their digital storefront, which is owned by the eBay stores team. This is going to continue to roll out that way across multiple different areas.
It’s not just about eBay using AI to build experiences, but it’s actually about putting AI into the hands of our sellers and buyers, so that — especially our sellers — can build experiences for their buyers. It’s actually the sellers using our technology to build 3D experiences or 3D visualizations of their products.
Let's talk a little more about that – about people who usually don’t work with AI using it, or even building it. There are lots of low-code and no-code AI tools out there. Are non-engineers, people on the business side internally at eBay, using low-code AI or auto ML tools?
We are putting that into the hands of our developers in some instances where we felt that as long as they operated under certain parameters, under certain constraints, they could scale it up independently without in-depth knowledge of AI or machine learning. We are doing that first in areas that are low-risk, where there’s not really an opportunity for bias or privacy issues, et cetera, no fraud or cyber issues.
That’s where we started, and we’ll proceed, but we have to be very careful as we do this because we need to understand what’s being put into production in front of our customers, and as you scale that up, you need all of the instrumentation in place to be able to continuously monitor.
With AI, the piece of software could be performing correctly, but you need to monitor it because the world changes and it’s reacting to the world.
Are there processes in place to protect against risks when others at eBay use some of these third-party low-code AI technologies?
I am standing up essentially standards, best practices and governance that includes membership and representation from across the company in order to ensure that regardless of the implementation, the same standards are being kept and monitored. I think one of the biggest challenges, one of the big differences between an AI solution in production and other software: Software needs to be maintained and monitored in general, but engineers that deploy a piece of code, a general piece of software, their need to maintain it really has to do with technical performance issues. It’s about whether the actual integration is still up to date.
With AI, the piece of software could be performing correctly, but you need to monitor it because the world changes and it’s reacting to the world. And data changes, the performance of the model changes, therefore you have to monitor it on an ongoing basis. That’s something that a lot of companies misstep there, where they treat it like software without really treating it like an ongoing — I don’t want to say living — but it’s something that is continuously changing and evolving and needs to be monitored.
So as we allow teams across the company to use no-code or low-code, and any kind of AI development, we need to have the right requirements and processes in place for ongoing monitoring.
What’s planned for 2022 when it comes to hiring on the AI team or other AI trends?
We’re not one of those companies that goes out and just gobbles up every person with AI in their title; we’re strategic about it. But there are specific areas — for example, computer vision and natural language processing — that are of strategic importance for us this year and so we’re going to focus on those.
There’s been tremendous progress in our ability to understand language in the terms that customers communicate to us in, so I think for us what you’ll see in 2022, you’ll see further developments in our customer assistance or customer service conversational capabilities, but you’ll also see more multi-linguality.
We’re going to be really double-downing on immersive experiences. So, 3D was sort of an early foray but you’ll be seeing more products rolling out, especially on our mobile platform, to close the gap between ecommerce and physical retail to enable buyers to experience products in an immersive way so they have full confidence in their buying decision. That really comes through computer vision, natural language processing, through personalization as well as some adjacent technologies like streaming, like computer graphics.