Chief AI & Data Officer Ting Cai unveils Rakuten AI strategy at Rakuten Optimism 2024
11 hours ago
Rakuten Chief AI & Data officer Ting Cai outlined the company’s latest developments in AI and shared his vision for the future of the technology in front of a packed business conference audience at Rakuten Optimism 2024.
AI and Omotenashi: Designing AI that proactively drives value
Cai opened the presentation by referencing a 2016 blog post where Rakuten Group Founder and CEO Mickey Mikitani stated, “I think we can all benefit from looking at how to use AI to deliver on the promise of omotenashi – the concept of achieving the height of possibilities in customer satisfaction, including the very anticipation of their wants and needs.”
This, Cai explained, is the core of Rakuten’s approach to AI. “We want to leverage AI to deliver the ultimate hospitality to our customers. Not only to meet their needs, but to anticipate what they want.”
Cai pointed to the recently released Stanford AI Index as a validation point for the opportunities in generative AI. The report estimates that global investment in generative AI grew to more than $25 billion in 2023, eight times the amount in 2022. Industry momentum is strong and growing stronger, and Rakuten is seizing the opportunities by leveraging Rakuten’s unique data, ubiquitous channels and operational excellence to utilize AI to augment human creativity.
Relentlessly pursuing innovation. Relentlessly pursuing safety.
He added that while Silicon Valley founders are known for encouraging a “move fast and break things” mentality, Rakuten takes a different approach – one that equally values a rapid pace of innovation and development, and an unwavering commitment to safety and trust.
This approach is deeply rooted in Rakuten’s AI Code of Ethics, a playbook that that sets guidelines on how the company can design, develop, test and deploy AI in ways that permit rapid innovation and experimentation, without ever compromising user trust or safety.
Cai then outlined the Rakuten AI Foundation and explained how by focusing on the fundamentals, like Rakuten’s unique data assets, powerful deep learning capabilities, and generative AI innovations, the company is well-positioned to scale AI to every part of the business. “We use our deep learning foundation to truly understand customer intent and personalize our services to their needs,” said Cai.
Cai dove deep into the underlying technology that’s powering Rakuten’s AI charge, starting with the concept of “embedding.” Embedding captures the meaning of words in mathematical terms, clustering similar entities together in the vector space, to help understand what a user means, not just what they say.
During the presentation, he cited an example where a customer comes to the Rakuten Ichiba online marketplace searching for an umbrella that provides “sun protection.” Embedding within the system will understand that this term is similar in meaning to “UV protection.” This association enhances the customer’s search, better matches them with relevant products and is now used with hundreds of millions of products across the Rakuten Ecosystem.
New similarities can be inferred using deep learning. To illustrate this point, Cai showed an example of how prescription glasses are categorized as “accessories” in the store but are also closely related to contact lenses, which are grouped in the “pharmaceuticals” category.
“Such discoveries of new relationships enable us to recommend related products to our users,” added Cai.
Deep learning is used to build applications for recommendations, search and ads in the Rakuten Ecosystem. Semantic search, which leverages deep learning, was launched this year with Rakuten Ichiba and last year with Rakuten Fashion, and the result thus far has been remarkable. Zero hit rate of customers (customers who search for products but cannot find any results), was reduced by almost 98 percent.
Rakuten is now extending this semantic understanding of words to visual search. Like word search, visual search uses transformer technology to break down images into patches and simulate them with words to predict the whole image. The Rakuten Ichiba mobile app uses this to allow customers to find more relevant search results from images they upload.
Improving customer experience with Rakuten AI for Rakuten Rewards
Following Ting on stage was Daniel Kellogg, Director of Analytics at Rakuten Rewards, who demonstrated Rakuten Rewards’ AI-powered solutions to address its most common customer service issue: discrepancies in Cash Back payments. Take for example a situation where a customer who bought a pair of sunglasses received a smaller Cash Back reward than they expected. Before AI, providing personalized answers instantly was costly and hard to deliver at scale. It required a team of highly-trained support staff who had to research the transaction, refer to a rule-based playbook, and then apply human judgement to the situation.
Rakuten Rewards is in the process of introducing an AI-powered workflow to answer inquiries of this nature instantly with messages tailored to the customer’s needs. Kellogg walked through a scenario where a customer did not receive the Cash Back they expected on a transaction. He started by entering the merchant name, date of purchase, expected Cash Back amount and order confirmation evidence in Rakuten Rewards customer service system to generate a claim ID.
From there, Rakuten Rewards’ support system generates an appropriate response based on factors like the customer’s purchase history, location and tone of their correspondence. After the order’s legitimacy has been analyzed and confirmed, “The AI system is intelligent enough to generate a member response and add the Cash Back amount to their Rakuten account,” shares Kellogg.
Kellogg expects this solution to resolve up to 90 percent of Rakuten Rewards’ customer inquiries, enabling its customer service personnel to dedicate their time to less-common issues that require direct human attention to solve.
“It delivers an experience that secures and deepens our customers’ trust and loyalty, while also reducing our operational costs by 90 percent,” he added.
Introducing Rakuten AI Universal Concierge
Cai and Zoey Zhao, a Rakuten AI product manager, unveiled Rakuten AI Universal Concierge – an experimental feature that enhances how customers experience the Rakuten Ecosystem.
Zhao explained that the Rakuten AI concierge provides a simple, unified way to look for products and services. She first demonstrated multimodal search with the Universal Concierge, using the example of a birthday gift search for a pet dog on Rakuten Ichiba. She uploaded a picture of the dog and provided a prompt asking if it had any gift recommendations. The application recognized that the dog was a poodle and provided ideas. Zhao then used her voice to ask the application for additional ideas on suitable raincoats for a dog. The speech from her mobile phone was recognized and resulted in further matching of dog raincoat products, with the application automatically listing suitable products.
This feature will be rolling out to select Rakuten services soon.
“By delivering personalized experiences to our customers, we increase our customer trust, earn their loyalty and deliver more business value to our merchants,” said Cai as he returned to the stage for closing remarks.
While Rakuten AI promises to continue bringing value to customers, Cai was clear that Rakuten AI is more than just that. “I believe Rakuten AI is a culture,” pointing to the culture of AI empowerment that defines Rakuten’s vision. “In the future we envision that AI will be a part of everyone’s lives… empowering people to accomplish more.”
https://rakuten.today/blog/...-strategy-at-rakuten-optimism-2024.html
|
Thread abonnieren
