After Meta, major tech giants are looking forward to their "Muse Moment"

Muse is opening a new narrative of "AI ROI".

In September this year, Meta launched the personal intelligent agent Muse, aiming to automate a series of tasks for consumers, including shopping, purchasing movie tickets, making appointment schedules and more.

At Meta's latest Q3 earnings call, CEO Mark Zuckerberg stated that Muse will advance commercialization earlier than the company's previous new products. In addition to the existing paid subscriptions, he also plans to collect fees from transactions facilitated by Muse, which are expected to be paid by merchants. Zuckerberg even predicts that Muse will eventually serve "billions of users", and the potential revenue generated by these users may exceed Meta's average revenue per user level.

Of course, Muse's current revenue volume is still small, far from being enough to drive Meta's overall revenue, but its greatest value lies in binding AI computing power investment with traceable and settleable transaction results for the first time.

Before that, the market could hardly answer a question: if a user spends an hour chatting with AI, how much is this duration worth? Muse gives the answer: AI helps users complete price comparison, reservation, procurement and service ordering, and the value is realized after the transaction is completed. This logic just provides a comparable reference system for domestic leading AI enterprises that are in a confused period.

While the other side of the ocean uses Muse to verify "Agent-driven transactions", the domestic AI track is simultaneously undergoing a quiet contraction and restructuring.

Since the beginning of this year, shutting down, merging, downgrading and converging entrances have become the main theme of AI products of domestic leading enterprises. It is not that AI demand has disappeared, but that the Agent 1.0 experiment of unlimited cost input and widespread deployment has officially come to an end.

While adjusting and launching new products, every leading enterprise may be looking forward to its own "Muse moment" in the future, because it means explosive product popularity, user growth, successful commercialization and rising company stock price.

Muse Commercialization Demonstration: Not Selling Ads, Selling "Deal Completion"

Less than two months after its launch, Meta has been talking at length about Muse's monetization strategy and prospects, and even plans to promote commercialization earlier than the company's previous new products.

Muse does have its own confidence. According to management disclosure, the most common usage scenario of Muse is shopping, more than a quarter of users complete purchases through it, with an average order value of more than 200 US dollars; the transaction volume in the first week of launch exceeded internal expectations, and the company needs to cooperate with Stripe to increase the credit limit. About 25% of users who have set up Muse use it every day, and the retention level is more than twice that of Threads in its initial stage. The company estimates that at least 50 million small and micro enterprises acquire customers through Meta's applications. Muse can serve both merchant customer acquisition and user transaction completion, and the commercialization path has taken initial shape.

According to previous statements, Muse will adhere to the ad-free model, ordinary users can use it for free, but advanced functions and enterprise services will be charged. Meta will make profits through multiple channels: transaction handling fees paid by merchants, subscription fees paid by heavy users and customers who want to use the most advanced models, and paid enterprise tools that rely on Muse for work.

At present, Meta has not announced the commission ratio, applicable merchant scope and official launch time. The final revenue scale will depend on Muse's task execution capability, merchant access status, and whether users are willing to hand over their shopping decisions to the intelligent agent. In the future, Meta will not only develop more types of Muse, but also connect Muse to more devices, such as smart glasses and the dedicated device Charm, to drive more people to keep paying through multiple touchpoints and entrances.

Zuckerberg said on the earnings call: "One of the reasons we don't charge everyone directly in advance is that we want to build the Muse network. We believe that when a large number of users and enterprises are using Muse, its scaled value will be higher." He added: "If computing power efficiency continues to improve rapidly in the next few years - which we believe is highly likely - Muse will bring substantial profits for many years to come."

Judging from this earnings report, most of Meta's revenue still comes from advertising. But Meta's Chief Financial Officer Susan Li predicts that the company's AI business will start to contribute incremental revenue from 2027, and will become "an important source of revenue diversification" over time.

Compared with the past two years, most of the domestic intelligent agent platforms of leading enterprises are still stuck outside the inherent mode of "creation - distribution - no one pays". Muse provides a consumer-oriented revenue stream that can truly reflect the output of large-scale investment. Even if the commercialization starts at a very small scale, the significance of Muse lies in establishing a measurable causal chain for AI investment: AI investment → user behavior change → transaction commission → revenue growth. A series of indicators such as millions of subscribing users, tens of billions of dollars in total merchandise transaction volume, app downloads, and user activity can all be used to measure the growth of the consumer AI business.

This is also why Meta's stock price was boosted after the earnings report, to a large extent because investors saw the prototype of this chain. The market no longer only cares about how much money it spends to build data centers, but begins to pay for the incremental revenue brought by AI applications.

Domestic Leading Enterprises' "Decluttering": Shutting Down Platforms, Betting on New Entrances

Compared with the vigorous progress of Muse, domestic AI products are undergoing another round of changes.

In the same week when Meta announced Muse's commercialization, Tencent Yuanqi team issued an announcement that the platform will stop all public services on November 9. Previously, Tencent QClaw has announced that it will cease operations at the end of the year. The creation entrance of Baidu Wenxin Intelligent Agent Platform was also closed in August. ByteDance took earlier and more thorough actions: in July, the Feishu product team merged with the Doubao team, the head of Feishu reports to the head of Doubao, and the commercialization team was merged into the Volcano Engine system. The AI short drama product Little Skylark was recently reported to be merged into Jimeng. Alibaba Cloud also officially announced recently that its virtual digital human business will be delisted.

The logic behind this round of shutdown, merger and restructuring of AI products is not difficult to understand, mainly for three reasons: First, compliance pressure. In July 2026, the *Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interactive Services* was officially implemented, and open UGC intelligent agents were included in strict supervision. Second, cost pressure. Independent products mean repeated investment in computing power, R&D and operation resources. Many products only get user stay data, but no revenue, so it is difficult to sustain before commercialization is realized. Third, strategic focus. In the past, leading enterprises collectively launched intelligent agent platforms, no-code tools and independent Agent applications, and a large number of products solved similar demands. When the capabilities of basic models are rapidly leveled, thin packaging layer products are easily replaced by the native capabilities of main Apps. Now leading enterprises are converging scattered capabilities to the core entrance to avoid "multi-front operations".

While doing "subtraction", leading enterprises are also doing "addition" - betting on personal intelligent agents. According to media reports, in the layout of personal intelligent agents, ByteDance's "Spell" project is led by the Doubao mobile assistant team, which has been in internal beta since April, and is planned to be launched after deep integration with Doubao; Tencent launched LightVela, and WeChat AI assistant "Xiaowei" has started internal beta; Alibaba Qwen App has officially launched "Qwen Go!", trying to handle flight claim settlement, shopping price comparison, telephone reservation and other affairs; Ant Group's "Lingguang" is also conducting small-scale internal testing of personal intelligent agents.

Previously, most domestic personal intelligent agents stayed at the vague positioning of "universal assistant", wanting to serve C-end users and take into account the B-end ecosystem at the same time, but often ended up achieving nothing on both sides.

The success of Muse also gives these new domestic personal intelligent agents a clearer reference, and establishes a new product development logic, that is, no longer building a generalized open intelligent agent square, but binding clear user groups and fixed high-frequency scenarios, embedding the commercialization path from the very beginning of design, with the goal of seizing the next-generation AI entrance at the mobile phone, system and hardware layer.

The Divide of AI Commercialization: Who Can Really Make Money?

Since the beginning of this year, an obvious signal shift is that the market's patience with AI has run out. Investors no longer pay for "doubling parameters" or "ranking first", they want to see revenue.

This is also why domestic leading enterprises have intensively charged for their AI products this year. In essence, it is the inflection point of shifting from "free for scale" to "charging by value". But what really deserves attention is not the "charging" itself, but that the charging logic is shifting from selling Tokens and selling memberships to selling results and selling closed transaction loops.

The research report of China International Capital Corporation also points out that the value of the C-end ChatBot model is relatively limited, and the space of Agentic AI is broader; in the long run, the C-end will most likely adopt the "free + performance advertising" model, and the performance advertising in the Agent era will be more transaction-oriented (CPS model).

At present, domestic AI commercialization has gradually shown a multi-level and diversified pattern:

1. In terms of MaaS/API services, leading manufacturers charge by call volume or Token, which is the most stable source of revenue at present;

2. In terms of C-end subscriptions, Doubao, Wenxin, Qwen, Kimi and other products have successively launched paid memberships, using higher context, faster reasoning, and exclusive Agent quota in exchange for stable cash flow;

3. In terms of pay-by-result (RaaS), customers pay for actual delivered quantifiable business results, for example, e-commerce marketing takes 5%-15% share by GMV, and financial risk control takes 10%-20% share by the reduced loss amount;

4. In terms of ecosystem sharing, the platform draws 15%-30% share from developers or transaction revenue;

5. Overseas commercialization: Domestic models accelerate going global with cost advantages, providing services for overseas B-end, developers and C-end users, using overseas payment habits to make up for the short board of low C-end payment conversion rate in China, becoming an incremental track. For example, DeepSeek and Zhipu AI have successively accessed the settlement systems of global top platforms such as AWS.

Returning to Muse's commercialization, the core of its monetization does not lie in subscription fees, but in the fact that it proves that AI can cut into the transaction chain and take a share from it. Before that, almost no domestic leading enterprise publicly disclosed the scaled revenue of similar models.

At the time of the second quarterly report this year, when Tencent mentioned Xiaowei, it once mentioned how to build a transaction closed loop between intelligent agents, that is, in the future, users only need to give a complex instruction to the exclusive intelligent agent, and the intelligent agent can automatically complete the entire transaction process. But Tencent's statement on this is also relatively cautious, emphasizing "user experience" and "privacy" rather than the monetization rhythm.

If we compare the commercial paths of intelligent agents at home and abroad, overseas Muse may be easier to run through the transaction closed loop of independent Agent first, while China will rely on the existing e-commerce, local life and enterprise service ecology, pilot the commission mode in small steps, use subscriptions as the bottom support, and gradually turn AI from a "chat tool" into a trusted digital agent.

No matter what the process is, the AI commercialization companies that can finally stand out need to meet three conditions at the same time: controllable computing power cost, stable task delivery capability, and value link that can be confirmed and attributed. Relying only on membership subscriptions or only on API calls will have a ceiling in the long run; players who can run through the combination of "subscription bottom support + result sharing + B-end productivity solution" will get the pricing power of the next-generation AI entrance.

Conclusion

Before last year, the narrative of the global AI industry was "input - catch up - re-input". Since the beginning of this year, the narrative has become "who is the first to find a billable closed loop". Meta gives an answer with Muse: let AI become the intermediary of transactions, not the porter of content. Merchants' willingness to pay for deal completion is far stronger than their willingness to pay for exposure. This logic holds true in the advertising era, and will only be stronger in the intelligent agent era.

Domestic leading enterprises are undergoing the transformation from "spreading tracks" to "calculating ROI". Shutting down inefficient products, converging core capabilities, and betting on personal intelligent agents, in essence, is to invest limited resources in fields with clearer commercialization paths. In the next stage, whoever takes the lead in running through the combined model of "subscription + transaction commission + ecosystem share" will take the lead in the competition for AI entrances. And computing power efficiency, compliance capability and ecological closed loop will become the three key variables that determine the outcome.

This article is from the WeChat Official Account "New Business Research Society", author: Shang Jing, published with authorization from 36Kr.