News Reporter |
News Editor | Wen Shuqi
When AI phones evolve from “being chatty” to “getting things done,” the relationship between phone makers and app developers is undergoing a major shake-up.
Recently, OPPO and Alipay teamed up to launch a multi-agent interaction solution. Users just chat with OPPO’s Xiao Bu assistant, and it quickly triggers Alipay’s agent “Abao,” tapping into a service network covering daily needs like food, housing, travel, and even government services. Right now, they’ve linked up nearly 200 common functions and rolled out over 18 digital public service scenarios.
This isn’t OPPO’s first move in the AI ecosystem. Over the past year, they’ve launched features like “One-Tap Ask” and “One-Tap Quick Save,” pioneered agent collaborations with WeChat for sending messages and making video calls, and publicly introduced the A2A (Agent-to-Agent) protocol.
For Jiang Yuchen, Director of OPPO ColorOS Smart Product R&D, these steps point to one clear judgment: shifting from apps to agents is a paradigm shift of the era, and it won’t change just because any single company wants it to.
In the cutthroat AI phone race of 2026, what’s the core value for handset makers in the agent ecosystem? To dig into these questions, NUPIAO recently sat down with Jiang for a chat.
Below is a lightly edited transcript of our conversation with Jiang:
NUPIAO: What shape should the so-called agent ecosystem take, and what role does it play inside a phone?
Jiang Yuchen: Honestly, I think the phone’s position hasn’t changed at its core. The OS’s niche has always been “aggregation and distribution.” The phone is the first layer of interaction between users and the digital world. In the past, it was more about distributing “app icons”—clicking on the desktop, push notifications, even our “app suggestions.” Going forward, it might be more about interacting through A2A, but the niche hasn’t shifted.
Future phone interactions will be more personal, smarter, and more precise. The phone will fully grasp user intent and distribute it to the most accurate service in a caring, spot-on way. It’ll let users accomplish complex, tedious daily tasks with minimal words—or even no words at all.
For example, a user says “buy me a ticket to Shanghai.” No need to clarify or click—it just books the best flight that matches my schedule and preferences. Going a step further, the user doesn’t even have to speak. The phone pops up a little card on the home screen saying, “I see you have a meeting scheduled in Shanghai. I noticed you haven’t booked flights or hotels yet. I searched X flights and hotels and picked the July 20th 8 PM flight on XXX, plus the Pudong XXX hotel. You’ll leave from Huarun Qianhai Building around X, arrive at the hotel by Y, and it’s a 38-minute ride to the venue the next day. Wake up at 8 AM. Want me to book it?” That’s personalized, proactive intelligence.
NUPIAO: What’s the real value of this iteration for users right now?
Jiang Yuchen: For users, it’s mostly about the translation layer. The phone has always been a translation layer between users and service providers.
Let me give you an example. Say, “buy me a ticket to Shanghai.” If you shoot that request directly to Meituan or any ticket vendor, they’d likely ask for the time, company, and price. What we do is “personalized memory.” Through the query, we understand the user’s schedule and preferences, do some reasoning, package that info into a refined query, and pass it to the right service provider. So, if it knows I have a meeting in Shanghai on the morning of July 21st, that I hate waking up early, and that I’m a China Southern Airlines member, it naturally searches for flights on the evening of July 20th, prioritizes China Southern, and sends that query to the service.
From the user’s perspective, I just said one sentence—”buy me a ticket to Shanghai”—or even said nothing, and the phone handled it. It cuts down on user interactions and the hassle.
NUPIAO: OPPO just announced a partnership with Alipay’s “Abao.” How is this symbiotic AI interaction different from traditional app switching?
Jiang Yuchen: In terms of what we can pull off, a typical query is “help me collect energy in Alipay.” That’s gotten a lot of positive buzz online. Before, you’d have to open Alipay, go into Ant Forest, and tap. Now, one sentence does it. Another example is topping up your phone. You’d normally open Alipay, select, and pay. Now, just say “top up my phone with 100 bucks on Alipay,” and it’s done. That’s a big shift in the user interaction paradigm.
NUPIAO: From a technical chain perspective, what technical adjustments did OPPO and Ant Group each make, and how are responsibilities divided?
Jiang Yuchen: We’re strategic partners with Alipay. We want to bring this “intent-to-service” experience to everyday users through our collaboration. We handle “intent understanding,” which includes memory and personalization. Alipay handles “service fulfillment,” actually closing the loop on the service.
NUPIAO: In recent talks with software developers and other manufacturers, I noticed an interesting trend. Unlike major internet players, many mid-tier companies are much more proactive about opening up their interfaces.
Jiang Yuchen: Different companies face different environments, and you can indeed see some gradually opening up. The shift from apps to agents is a paradigm shift of the era. It won’t be swayed by any single company’s will. Last year, the agent ecosystem might have been 0% ready. This year, it’s maybe 10% to 20%. It’s an exponential curve. I believe by the year after next, or the one after that, it could cover 80% to 90% of the ecosystem.
NUPIAO: The big model capabilities are mostly in the hands of model providers. What’s the value of phone makers in the agent ecosystem? What’s OPPO’s core competitive strength?
Jiang Yuchen: Our view has been consistent and clear: it’s all about memory. We believe memory is crucial in the entire AI ecosystem. The flight booking example I gave earlier—it combines schedule, preferences, and membership info for deep reasoning, delivering a hassle-free experience. That’s something the model layer can’t and shouldn’t do. Phone makers are the devices that stick with users the longest. A “cross-app” comprehensive user memory, built on user authorization, will be our core competitive moat.
Memory has three layers.
First, it’s perception of the user’s time and space dimensions. With full user consent, we sense usage time, location, sound, visuals, etc. Once we gather enough data, it becomes valuable.
Second, it’s understanding the data. Connecting different data points allows us to model and understand a person.
Third, it’s self-evolution. The memory we build isn’t set in stone. For instance, you might love Taylor Swift today, but tomorrow you could be more into Justin Bieber. Preferences change over time. I think self-evolution isn’t just about adding information; it’s also about forgetting. Your brain can’t remember everything. A lot of memories fade and blur over time.
Each of these three layers—perception, data understanding, and self-evolution—is tough to nail. So this is a technology we’ll likely spend a long time building.
NUPIAO: Phone makers are all trying to build new business models with AI. Where will OPPO focus? Will these models converge?
Jiang Yuchen: OPPO isn’t rushing into AI commercialization right now. We’re more focused on building an AI ecosystem. Over the long haul, the essence of business models often doesn’t change. The OS’s niche has always been aggregation and distribution. I still think in the AI era, the way to close the business loop is aggregation and distribution. From that angle, future business models are pretty easy to build. Our attitude is, we’re in no hurry.
NUPIAO: The API route has made big strides this year, but the reality is that major internet firms haven’t fully opened up deep scenarios and data. Is the industry waiting for a revolution?
Jiang Yuchen: Opening deep APIs doesn’t make business sense. From a technical standpoint, the best approach is still A2A. That way, everyone can have a relatively standard framework, making it easier for the ecosystem to thrive.
NUPIAO: Privacy and security are core to the system. On a practical level, does OPPO’s memory system have any interception mechanisms in place?
Jiang Yuchen: User security and privacy are our top priority. It’s part of our social responsibility as a large-scale phone maker. We’re definitely investing more. Some of our moves might seem “conservative,” but that’s actually being responsible for user security and privacy.
Technically, first, I’ve stressed the superiority of A2A. It’s not just easier to promote ecologically; it’s also safer. All interactions happen at the agent layer, without touching underlying data or high-risk operations. Permission controls and sensitive info exchanges during execution require user confirmation. Especially when money’s involved, we’re extra cautious. Without user authorization and confirmation, we never execute operations.
We’ve built many security measures into the “memory” tech chain. First, on the device side, we have a fully sandboxed runtime environment. Your data stays on your phone, and other apps can’t access highly sensitive info. Some data has to go to the cloud, and we use PCC (Private Cloud Compute) there. That means even on the cloud, OPPO can’t see your data, others can’t, and you can’t either. It’s computed in a completely private environment.
Even when data goes to the PCC cloud environment, we do desensitization calculations for highly sensitive info. For example, if you say “find my bank card password,” when it’s uploaded to the cloud, it replaces the password with a random string of characters—including your name. When the model does its reasoning, the algorithm can’t access the real info. These three layers of protection keep user security and privacy in a relatively safe environment.
When it comes to sensitive permissions, we require user confirmation. Our attitude is especially cautious.
NUPIAO: How do you ensure user awareness?
Jiang Yuchen: We always get user authorization now. All data and permissions require user consent. You can fine-tune each item in the settings.
NUPIAO: Some internet firms argue that on-device vs. cloud computing is a false dichotomy. They say user data already lives in the cloud, and from a hardware cost logic, phone makers emphasizing on-device cloud privacy is partly a way to shift costs?
Jiang Yuchen: I think privacy is more important than cost. It’s closely tied to user personalization. Our product philosophy is called “Personalized Proactive Intelligence.” It’s about understanding and modeling users. As “Personalized Proactive Intelligence” becomes mainstream, on-device computing will become even more vital. I believe on-device processing is still very necessary. The most complete information is best kept on the device—that’s the safest.
NUPIAO: Between democratizing AI capabilities and hardware performance limits, how does OPPO ensure “digital fairness” for users across different price points?
Jiang Yuchen: Our entire tech architecture is device-cloud collaborative. Overall, simpler tasks or ones that need long-term processing are handled on the device. We fully leverage idle times—like when you’re sleeping or charging—to minimize power and memory impact. For users without on-device models, we try to keep the feature experience consistent for everyone. It’s just that with an on-device model, some features get a boost.
