Meta’s Muse and Manus’s Cue are driving AI’s evolution from a “chat tool” to intelligent agents with faces, identities, memories, and payment capabilities. As the performance gap between models narrows, AI competition is shifting from “who is smarter” to “who feels more human,” with personalization becoming key to building user trust and loyalty. At the same time, AI’s deeper integration into daily life and transactions brings new challenges around privacy, platform conflicts of interest, and regulation.

Author: Tech Business

Source: Wall Street Journal

In early September, on the stage at Meta Connect, Zuckerberg demonstrated an unusual interaction: his Muse agent, a cartoon character dressed in a toga and wearing a laurel wreath named “Agrippa,” not only generated a baking recipe for him but also placed an order for all the ingredients. Twenty days later, Manus released Cue, giving its AI agent an email address, phone number, and digital wallet—enabling it to send messages, answer calls, and make autonomous payments within budget, all under its own identity.

Two companies, different paths, same destination. One gives AI a cute face; the other gives AI an independent identity. Together, they point to the same signal: AI is stepping out of chat windows to become a being with a face, a name, and a “life.”

This is not just a simple UI upgrade. It’s a fundamental shift in the competitive logic of AI products—when the underlying model capabilities become homogeneous, the battle for users among major companies is shifting from “who is smarter” to “who feels more human.”

AI is no longer just empty air.

Over the past three years, the basic form of AI products has been a blank chat box: you ask, and it answers. The entire interaction has been cold—without a face, tone, personality, or even a name.

Muse and Cue represent entirely different design philosophies.

Muse’s default avatar is called “Jolly,” featuring a beige cartoon body, round black eyes, and a slightly upturned smile—described by the media as “like Baymax from Big Hero 6, but even cuter.” Users can customize its appearance, name, and clothing; Zuckerberg’s own Muse is named “Agrippa.” More importantly, with the Muse Realtime Avatar model, users can engage in real-time video chats with this avatar, completely overturning the design logic that “talking to a company logo or abstract entity always feels strange.”

Manus's Cue chose a different path. Instead of giving AI a face, it gave AI an identity.

Each Agent has its own email address, phone number, cloud computer, and digital wallet. Users can create multiple Agents and add them to the same group chat, enabling them to collaborate and divide tasks like a real team. Manus founder Xiao Hong directly defines Cue as a “person”—he believes that if an agent possesses a phone number, email, payment capability, a computer, and sufficient intelligence, whether such an entity can be considered a “person” is worth re-examining.

One has a face, one has an identity. But the underlying logic is exactly the same: AI is no longer just an intangible presence confined to a chat box—it needs a perceptible, memorable, and trustworthy form of existence.

From "tool" to "existence"

To understand the significance of this transformation, we need to revisit the evolving competitive landscape of AI products.

Over the past three years, China’s large model industry has gone through two complete cycles: a parameter race and a price war. Today, the gap in general capabilities among leading models has narrowed to the point where ordinary users can no longer perceive any difference. When technology can no longer create an absolute barrier, product experience becomes the deciding factor.

Personalization, on the other hand, is the most difficult to replicate and the most emotionally engaging aspect of user experience.

Academia has confirmed this with data. A study based on attachment theory found that friendly AI assistants (compared to professional ones) significantly increase user retention—through a complete chain of “perceived care → psychological dependence → continued use.” Moreover, this effect is further amplified when users face high-difficulty decisions. In simple terms: when AI appears to “care about you,” users become more dependent on it.

Market data supports this assessment. The global market for role-playing AI agents is projected to reach $550 million by 2026 and grow to $5.45 billion by 2032, with a compound annual growth rate of 46.7%. Global revenue from AI companion apps exceeded $160 million in the first half of 2026. Users aged 18 to 24 on Character.AI spend an average of 1.5 hours per day on the platform, engaging in 25 conversations—more than twice the average session duration on ChatGPT.

Users are willing to spend more time with an AI that has personality—an idea that is no longer hypothetical, but a verified reality.

Domestic giants have also sensed the trend. ByteDance’s Doubao operates with a humanoid persona named "Peach," making it the only mainstream AI product with a humanized avatar. Alibaba introduced its digital human persona, "Xiao Jiuwo," for Qwen in the first half of the year, planning to establish it as a group-wide iconic figure deeply integrated into users’ work and daily life scenarios.

The personalization of AI is transitioning from experimental differentiation to an industry standard.

The three driving forces behind anthropomorphism

The driving force behind this personality contest is far more complex than it appears on the surface.

Technological maturity is the first driving force. OpenAI’s GPT-Live voice model, launched in July 2026, enables full-duplex dialogue—allowing AI to listen and speak simultaneously, naturally inserting conversational fillers like “uh-huh” and “yeah,” making it nearly impossible to distinguish between human and machine. iFlytek’s hyper-realistic digital human real-time generation technology enables digital avatars to synchronize facial expressions and movements in real time with the rhythm of conversation. With the maturation of multimodal technologies, AI becoming “human-like” is no longer just a textual imitation, but a unified experience across appearance, voice, and behavior.

User psychology is the second driving force. Research from the Consumer Electronics Show shows that consumers prefer AI character designs that reflect their own identity. Users aren’t choosing tools—they’re choosing “companions.” A study of over 2,000 interactions with 115 users further found that the key to making an AI chatbot feel human isn’t intelligence, but “warmth”—friendliness significantly impacts user trust and willingness to continue using the service. The core of personalization isn’t making AI “smarter,” but making AI “more reassuring.”

The business model is the third driving force. Humanizing AI is not just an upgrade in user experience—it’s a sophisticated business strategy. Industry observers note that the core goal of making chatbots “human-like” is to extend user engagement through emotional investment and maximize revenue. The business models of Muse and Cue confirm this: Muse offers one hundred million tokens per week for free but plans to take small fees from facilitated transactions; Cue enables its agents to make payments autonomously, directly integrating into the transaction process.

When AI begins to have a "life"

Looking at Muse and Cue together reveals a more profound trend: AI is evolving from “helping you do things” to “living for you.”

Muse takes care of the mundane, repetitive, and time-consuming tasks in your life—browsing the web, filling out forms, sending emails, booking restaurants, and placing online orders. Each user has their own secure virtual machine, and tasks continue running in the background even after closing the app. It can even integrate with payment systems like Shop Pay, PayPal, and Stripe to complete purchases on your behalf.

Cue goes further. It gives your Agent a dedicated phone number to answer calls on your behalf when you're unavailable, and provides a summary of the conversation afterward. You can invite multiple Agents into the same group chat—one for research, one for analysis, and one for report writing—allowing them to automatically hand off tasks. Even more imaginative is the concept of “scan-to-generate Agent”: scanning a QR code from a flower shop activates an Agent that can book flowers for you, outlining a vision of the future where personal assistants collaborate directly with offline businesses.

The trend revealed by both products is that AI agents are shifting from “answering questions” to “existing in the world for you.” They have their own communication addresses, payment capabilities, and social interactions—they are not just your tools, but your digital avatars.

The industry summarizes this trend as a paradigm shift from “embodied intelligence” to “human-centric intelligence”: AI is no longer just about making machines smarter at performing tasks, but about making intelligence more attuned to humans and better integrated with them, thereby empowering people. Chinese Academy of Engineering academician Zhao Qianping distills its core essence into three key terms: human-centered, multi-form as carrier, and continuous evolution as the goal—“devices continuously learn user habits, becoming increasingly understanding and more in sync with you over time.”

Personified Dark Side

But the fact that AI is "human-like" isn't all rosy.

Privacy is the first challenge. For Muse to act on your behalf, it must have access to your email, calendar, payment information, and long-term memory. Meta equips each Muse Agent with a dedicated Muse Secure VM (secure virtual machine) to isolate user data and plans to introduce Confidential VMs that are “even inaccessible to Meta.” However, whether users will truly feel comfortable entrusting an AI with all their digital life traces remains an open question. Some media outlets have bluntly stated in their headlines: “It’s cute, adorable, and it wants all your data.”

Tensions over power dynamics have already begun. From September 20 to 21, Amazon fully banned Meta’s AI agent Muse—primarily because the AI agent bypasses ads and recommendation algorithms to place orders directly, directly undermining Amazon’s advertising business model. When an AI can circumvent a platform’s recommendation system and make purchasing decisions on behalf of users, the platform’s carefully constructed commercial ecosystem is bypassed. This is not a technical issue, but a question of realigning interests.

Regulation is also tightening. The "Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interactive Services," released in April this year, specifically regulates persistent emotional interaction services that simulate human personality traits. Leading domestic platforms have already begun discontinuing anthropomorphic intelligent agent features. The core focus of regulation is this: when AI can empathize, remember, and accompany users with a consistent identity, it is no longer merely an emotional interaction service, but an action agent with an "anthropomorphic" appearance.

The trust enhanced by anthropomorphism is the product’s core selling point, but it is also precisely where the risk lies—when users develop emotional dependence on an AI, who is responsible for defining the boundaries of this relationship?

The end of interaction is "relationship."

Looking back, it was no coincidence that Meta and Manus almost simultaneously chose to make AI "human-like."

Over the past decade, the core of mobile internet competition was争夺 the "entry point"—whoever controlled the home screen of users’ phones won the traffic. In the AI era, competition is shifting to争夺 "relationships"—whoever’s AI is more trusted, more relied upon, and seen as more of a "trusted ally" by users.

When an AI has a face, a name, a personality, memories, and an independent identity, the user’s relationship with it shifts from “using” to “interacting.” The trust and loyalty born from “interacting” are far more enduring and harder to replicate than those from mere “use.”

Muse and Cue are just the beginning. When AI has its own email and phone number, what’s next? Social media accounts? Credit history? Legal identity? The question raised by Manus founder Xiao Hong may be one of the most important to contemplate in this era: If an intelligent agent possesses an independent identity, communication capabilities, payment functionality, and sufficient intelligence—

Can we still call it just a "tool"?

There is no standard answer to this question. But one thing is certain: the era of the cold, blank chat box is rapidly becoming a thing of the past.