Ant Group releases Lingguang 2.0, positioned as a personal AI assistant that can get things done

Double 11 is here, and for seasoned shoppers there is often a long list of preparations to complete before placing an order.

First make a list, compare prices across platforms, calculate coupons and spend-based discounts, then set an alarm for paying the final balance. After finally finishing the purchase, a couple of days later you find the price has dropped again, and you still have to check the price-protection rules, find the order, and apply for a refund of the difference.

In September this year, Meta launched Muse, which reached 5 million downloads within three weeks of going live. This type of Personal Agent demonstrates a new possibility: the user states a goal, and the AI keeps working after the conversation ends, contacting the user again when there are changes or when confirmation is needed.

On October 27, Ant Group released Lingguang 2.0, positioned as a "personal AI assistant that can manage things," taking the first baton among major domestic companies for Personal Agents. What it wants to take over is precisely those matters that require constant attention: remembering what the user has asked, continuously monitoring changes, reminding the user promptly at key moments, and helping the user move things forward step by step.

Why are Personal Agents appearing in concentrated fashion at this moment?

From Muse to Lingguang, different approaches to Personal Agents

From Muse to Lingguang, Personal Agents are exploring a new kind of user relationship: the user assigns a matter, and the AI keeps thinking about it on their behalf.

This change has a technical foundation. As models' ability to execute long tasks improves, AI is now able to continue working after the user leaves the conversation. At the same time, long-term memory, cloud-based operation, task scheduling, browser operation, and tool-connection capabilities have gradually matured, enabling AI to remember goals, monitor changes, and take action at different points in time. Personal life and affairs management have therefore become a new testing ground.

Compared with Agents for office work, programming, and similar fields, which mainly target knowledge workers or programmers, Personal Agents touch a broader range of user needs: shopping price comparison, monitoring prices for flights and hotels, managing bills and automatic renewals, arranging family trips, organizing information from class group chats and work group chats, and even pet care, vehicle inspections, and vaccine reminders.

These small matters of daily life share one thing in common: the matters themselves are not necessarily complex, but they need to be remembered over the long term, and the relevant information keeps changing. What users need is someone to keep watching on their behalf, to remind them promptly when there are changes, and to push things forward when action is needed.

When AI can search, compare, and continuously follow up around user goals, users' choices shift from "which app to open to find a service" to "which AI to hand the matter to." Competition among platforms and merchants will change accordingly: whether services can be discovered and invoked by Agents, and whether transactions and fulfillment can be completed. Personal Agents therefore have the opportunity to become a new service entry point and a key focus of global giants' layouts.

Muse places more emphasis on autonomous execution around goals. In the scenarios described in its official introduction, the AI can handle web pages, email, and online services, helping users send emails, book travel, and fill out forms, and can also continue advancing long-term goals after the user closes the app. When sensitive actions such as sending emails or making purchases are involved, Muse requires user approval.

This path is related to the overseas internet service environment. Overseas, email is the information hub through which people manage daily affairs, carrying information such as orders, bills, appointments, and itineraries; websites and online platforms are important entry points for obtaining services and completing transactions. Through browsers and connectors, Agents can obtain this information, access related services, and continuously advance tasks.

Lingguang 2.0 represents another path. It does not position itself as "doing everything for the user," but instead focuses on "managing things": starting from the daily needs of Chinese users in consumption, travel, family, and work, it lets AI remember matters, continuously follow up, and connect to services familiar to users such as Alipay, DingTalk, and Feishu, so that matters requiring repeated worry can be managed over the long term.

Understanding this path of Lingguang can be approached from three aspects.

From the perspective of the digital service environment, Chinese users' consumption, travel, work, and family affairs are often distributed across different apps, mini-programs, and platforms. Cross-platform information still needs to be organized by users themselves, and long-term matters still need to be remembered by users themselves. The opportunity for Personal Agents lies in taking over this work of continuous worry.

From the perspective of service connection paths, Lingguang 2.0 relies on Alipay's service ecosystem and Ant's payment and trust foundation, and connects to services such as DingTalk, Feishu, email, and Mijia through tool connections. This gives it the opportunity to connect scattered information and services around a single matter of the user, and to link together subsequent organization, reminders, and follow-up.

From the perspective of product experience, many domestic services are already convenient enough; what is more worth looking forward to is extending the convenience of a single operation into peace of mind for the whole matter. "Handling a task" is more like a one-off task: the user gives an instruction, and the AI completes the action; "managing a matter" means that after the task is completed, the AI still remembers the matter, continuously watches for changes, and reminds the user and advances the next step when needed.

Lingguang 2.0, from handling a task once to continuously managing matters

Turning the convenience of a single operation into the peace of mind of long-term entrustment is precisely the main thread of Lingguang's upgrade. The capability improvements of Lingguang 2.0 can be summarized in four points: it remembers, it can advance, it follows up, and it can adjust. The value of these capabilities needs to be understood in the context of concrete daily life.

First is remembering matters. For example, when arranging a family trip, the user says "bring the elderly, avoid early-morning flights, and choose a hotel close to the subway," and these conditions must carry through subsequent arrangements. When discussing flights and hotels a few days later, Lingguang can still continue from the previous requirements without the user having to explain from scratch each time.

Next is continuous attention to changes. For example, if there is a refrigerator on a shopping list and the user says, "Remind me when the price drops below 2,000 yuan," they no longer need to open the app every day to check the price. Lingguang keeps following up and reminds them again when the conditions are met. The user puts down the phone, but the matter is not dropped.

Further on, changes are turned into actionable items. For example, when a teacher posts a long string of spring outing notices in the class group, Lingguang can extract them into several clear to-dos: "Hand in the permission slip by Wednesday; bring a water bottle and raincoat on Friday." Users do not have to scroll through chat history repeatedly to know what to do next and when to do it.

When new situations arise, arrangements also need to be adjusted according to existing goals. For example, on a weekend family outing, if the weather forecast suddenly calls for rain, they need to consider switching outdoor activities to indoor ones and rescheduling travel times. Lingguang can help sort out adjustment plans; decisions involving cancellations, rebookings, and the like are then confirmed by the user. Continuously managing matters also includes helping users think through the next step when plans cannot keep up with changes.

This is precisely the "China solution" that Lingguang 2.0 offers for the Personal Agent: it does not package itself as an "all-capable errand assistant," but instead starts with memory, organization, monitoring, reminders, and follow-up, connecting the technical capability of continuously managing matters with China's mature digital service ecosystem, so that the life services users are familiar with gain an extra layer of more sustained care.

Ant Group steps up Personal AI to bring the service ecosystem to life

For the Personal Agent to become a new service entry point, in addition to model capabilities, it must pass through two gates.

The first is the gate of services. For an Agent to manage matters, it must enter the apps and services users use daily, but high-frequency applications in China generally set barriers to Agent access. What can be accessed and how far a task can be handled depend on the degree of service openness and authorization rules.

The other is the gate of trust. When AI begins reading bills, following up on matters over the long term, and even participating in payments, users need to be clear about what it can see, what it can do, which operations require confirmation, and how problems can be traced. Whether users are willing to entrust matters to it needs to be gradually established through clear permissions and reliable execution.

Ant Group's accumulated experience in these two areas provides the foundation for Lingguang 2.0's exploration. Relying on Alipay's service ecosystem, Lingguang can connect to services such as bills, memberships, and travel, and through tools connect to applications such as DingTalk, Feishu, email, and Mijia; Ant Group's 22 years of accumulated experience in payment security, identity authentication, and authorization management supports AI in handling matters within the scope of user authorization.

This is also the significance of Lingguang 2.0 as a window for technological exploration: it verifies capabilities such as long-term memory, task scheduling, and tool connectivity, and also observes which matters users are willing to hand over to AI and to what extent. Only when technology can run continuously and users are willing to entrust matters continuously can personal agents truly enter daily life.

At the AI payment level, Alipay has launched full-stack products such as AI Pay, AI Collect, AI Wallet, and Token Pay to support agents in completing transactions. In May this year, Alipay disclosed that AI payment already supports 95% of general agent frameworks, with a cumulative 300 million agent payments completed. Among them, AI Wallet supports users in managing agent tasks before and during payment, and querying bills after payment, enabling users to keep track of the process of AI spending money on their behalf.

At the AI application level, Ant Group explores Personal AI from different scenarios through Afu, Abao, and Lingguang. Afu focuses on health consultation and health management; Abao relies on Alipay to connect daily services such as consumption, travel, and errands; Lingguang focuses more on cutting-edge products and technological exploration, attempting to take on long-term matters that require continuous attention and follow-up.

It is worth noting that there is room for further collaboration between Lingguang and Abao. The product forms and technical capabilities that Lingguang explores first, once verified, are expected to be applied to Abao, reaching more users through Alipay's high-frequency scenarios and service ecosystem, allowing cutting-edge exploration to gradually move toward mass applications.

In addition, Ant Group is also extending AI capabilities to industry scenarios. For example, Ant Digital Technologies has built an "Agent Super Factory," helping industry partners develop and apply agents through platforms such as Agentar.

Simply put, only when AI payment, AI applications, and the ecosystem operate in better coordination can Lingguang "manage" matters better; and the continuous evolution of Personal AI like Lingguang is also expected to further activate the digital service ecosystem Ant Group has accumulated over the years, spawn new service demands, and open up greater room for imagination.

Perhaps, at the next Double 11, people can go to bed earlier while AI keeps watching prices; and when the class group chat hits 99+, they will not discover only before sleep that "a empty cardboard box is needed tomorrow."

For users, the value of Personal AI may be a little less last-minute panic: those "don't forget" items that always circle in the mind can gradually be handed over, matters are remembered for you by Lingguang, services always appear when needed, and life can be a little less frantic.