Dawn of the Super Agent: Mikitani at Rakuten AI Optimism 2026
18 hours ago
At the 2026 Rakuten AI Optimism conference, Rakuten Group Chairman and CEO Mickey Mikitani made it clear that the AI boom is not just another passing trend.
“We are entering an era where most people are using AI, even without realizing it,” he told the packed conference hall crowd. “The key question is, is this AI revolution the same as the internet revolution? And the answer is no, it’s totally different. It has an impact far beyond simply making things more efficient; it is changing the definition of almost everything.”
From Agent to Super Agent
Mikitani laid out Rakuten’s AI goals plainly and clearly: “We intend to become the platform that utilizes AI the most in the world.”
The integration of Rakuten AI in its services and systems is creating a seamless experience across the entire Rakuten Ecosystem – instead of isolated tools for businesses like Rakuten Ichiba, Rakuten Travel, or Rakuten Mobile, the company is bridging them all together into a unified Super Agent.
“Rakuten Group’s unique strength lies in our comprehensive ecosystem: shopping, travel, credit cards, golf, banking, securities – covering nearly every aspect of daily life,” Mikitani told the capacity audience. “By understanding all this connected data, the Super Agent acts as a personal executive assistant… From consultation to service booking and payment completion, our Super Agent works seamlessly across services.”
This high level of integration is allowing Rakuten AI to effectively function in the real world.
“The interface won’t be limited to the Rakuten Ichiba website or app; it can reach users through multiple channels – allowing you to shop via messaging alone,” Mikitani noted. “It’s also integrated with third party external apps and tools like maps. If a user from Tokyo travels to Sendai, it can suggest local specialty sweets on the spot.”
To help illustrate what this level of hyper-personalization offers, Rakuten’s Yupeng Zhang walked the audience through a live demonstration of the Super Agent planning a family trip to Okinawa. By tapping into the AI’s memory of past user behaviors, the agent prioritized accommodations with a kitchen and even added sunscreen to the order list, recalling that he had suffered a sunburn on his vacation the previous year.
Through simple natural language, Zhang then instructed the agent to adjust the trip for a “more relaxed pace,” whereupon it blocked the dates on Zhang’s calendar app, drafted an itinerary email to his wife, and even suggested tricks to boost his Rakuten Point multipliers.
“This Super Agent capability is a service unique to the Rakuten Ecosystem,” Mikitani noted. “While open architectures allow for many things, having rich historical data within an integrated ecosystem – alongside deep business relationships with merchants and hotel partners – allows us to achieve a far higher level of accuracy.”
Meet your new AI Store Manager, Ichiko
Rakuten’s vision for a Super Agent ecosystem ensures that humans remain firmly in the spotlight. Mikitani has long pushed back against the idea of e-commerce platforms functioning as faceless vending machines, and this philosophy has a place in Rakuten’s AI strategy, too.
“Ultimately, a real human handling things is best. At the same time, humans can’t process tens of thousands of requests 24/365,” he told the Optimism crowd. “By loading the store manager’s profile, personality, and persona, each store can have its own AI Shop Manager.”
Mikitani held an onstage demonstration of Rakuten’s new AI Store Manager, tentatively named Ichiko. Leveraging a merchant’s specific personality and deep product knowledge, shops will be able to create their own 24/7 AI avatars capable of handling complex inquiries and assisting customers with their shopping.
Instead of simply fetching generic product specifications from the public web, the AI Store Manager leverages Rakuten’s ecosystem to read a shopper’s broader ‘persona’ and tailor responses accordingly. A user shopping for golf clubs, for example, will be recommended an entirely different set depending on whether their history shows they are a seasoned player or a complete beginner.
“Humans need human treatment,” Mikitani stressed. “We need AI with a human touch.”
Taking AI in-house
To navigate the global AI revolution, Rakuten is leaning heavily on its proprietary data and developing its own Japanese-centric LLMs to avoid escalating costs and security concerns.
“In business terms, last year AI contributed about 25.5 billion yen in profit for the Rakuten Group,” Mikitani shared. “Why does Rakuten have an advantage over other companies? AI is ultimately a processing algorithm; it’s an equation. The variables in that equation are data. Data is the diamond in the rough, and the quality and volume of that data are crucial.”
Rakuten Group Executive Officer and Director of the AI Services Supervisory Department Taku Okoshi joined Mikitani onstage to shed light on Rakuten’s rapid deployment of these in-house models.
“At the beginning of this year, we released our 700-billion-parameter model, Rakuten AI 3.0. Looking at the Japanese AI market, several companies are building proprietary models, but as of now, we believe we are the only company developing a 700-billion-parameter class model,” Okoshi said, noting that the LLM accounted for roughly 19% of all AI token usage within Rakuten. “Beyond internal use, we plan to power the backend of our customer-facing services like Rakuten Ichiba and Rakuten Travel.”
The true power of Rakuten’s approach lies in the ability to leverage deep user histories without prohibitive computational costs. Yu Hirate of the Rakuten Institute of Technology gave a compelling demonstration of this novel architecture in action.
“We fed 18 years of purchase history of a single user (Hirate himself) into the LLM currently under development, and asked it: How have this user’s purchasing trends changed over time?” Hirate explained. “The LLM successfully detected that starting around 2021, I began repeatedly buying products from a specific brand. During COVID-19, as work-from-home increased, I bought various gadgets, fell in love with one specific manufacturer, and kept buying their products. The LLM accurately captured this behavioral shift from 18 years of data.”
Japan’s position in the AI race
“Looking at the bigger picture: Who will win the AI race?”
While a handful of frontier AI labs continue to push the pace, releasing industry-leading models and technology, Mikitani sees another significant trend in motion: the movement towards open-source models.
“Open-source AI will likely continue to rise and compete. Looking at American and Chinese [open source] models, they are catching up at amazing speeds. Therefore, it remains to be seen who will win. It’s even possible that AI models will transition completely to open source.”
As for Japan’s place in the AI race, Mikitani noted that the country “is lagging about 3, 4, or 5 laps behind.” However, he added, “we at Rakuten Group believe it is essential for us to develop our own fundamental AI [to empower Japan].” Many of Rakuten’s own AI models are based on open models, including its largest models like Rakuten AI 7B, Rakuten AI 2.0 and Rakuten AI 3.0.
Grounding AI in human empathy
Regardless of technological leaps or global competition, Mikitani stressed that AI development must remain grounded in fundamental human empathy.
“An AI acting as a shopping agent shouldn’t just execute transactions; it needs to understand intent, make proactive suggestions, and maintain the unique characteristics of each shop.”
For Rakuten, meanwhile, the future of AI isn’t about stripping away this humanity, but empowering it.
“AI is going to fundamentally evolve our services. But no matter how far we go, the most important thing is how we value the human element. We have always maintained that internet shopping is not a vending machine,” Mikitani concluded. “I believe that building this fundamentally Rakuten-style AI is what is important.”
rakuten.today/blog/...kitani-at-rakuten-ai-optimism-2026.html
Zurück
