Author｜Haotian

On September 14, 2026, Doubao officially rolled out the consumer edition of its mobile assistant, with the Nubia NaviX Ultra as the flagship model, slated for sale on September 16.

Compared to the technical preview version released at the end of 2025, the consumer edition of Doubao Mobile Assistant places a stronger emphasis on stability and security, yet retains the controversial 'mobile phone operation' feature. With this, Doubao Mobile Assistant can carry out complex tasks across various apps based on user instructions.

Owing to certain distinctions from traditional smartphones, the consumer edition of Doubao Mobile Assistant has drawn significant market interest. As of the time of reporting, the Nubia NaviX Ultra has racked up approximately 380,000 pre-orders on JD.com. According to Jiemian News, the total stock of the Nubia NaviX Ultra stands at around 200,000 units, hinting at a potential sell-out scenario upon its release.

However, even if all stock is depleted, a shipment volume of 200,000 units is relatively minor in the smartphone industry. As a product aiming to disrupt the industry, hundreds of thousands of sales are just the first step on a long road. Only by achieving blockbuster status like the iPhone and establishing a closed-loop commercialization model can Doubao Mobile Assistant potentially usher in a new era of AI smartphones.

01 GUI Controls Third-Party Apps: The Initial Doubao Mobile Phone Faces Ecosystem Barriers

In fact, the technical preview version of Doubao Mobile Assistant, previously released for developers, had already generated market buzz with its 'mobile phone operation' capability.

Image Source: Doubao

In December 2025, the Doubao team, in collaboration with ZTE, unveiled the Nubia M153 engineering prototype equipped with the technical preview version of Doubao Mobile Assistant. It could perform cross-app automatic operations through GUI simulation clicking technology, priced at 3,499 yuan. It sold out the day after its listing, with resale prices on second-hand platforms skyrocketing to tens of thousands of yuan (at one point exceeding 10,000 yuan).

One key reason Doubao did not immediately release the consumer edition after the engineering prototype's popularity was that the GUI Agent encountered permission and security constraints within the app ecosystem.

Image Source: Doubao

Rather than controlling third-party software through official interfaces, Doubao directly identified UI components via GUI, bypassing the App's traffic entry points, data sovereignty, and security risk control systems. After the launch of the technical preview version of Doubao Mobile Assistant, it swiftly faced restrictions from WeChat, Taobao, and financial apps, compelling Doubao to disable automatic operation capabilities for some apps.

In response, at the 2025 annual employee conference, Tencent founder, Chairman, and CEO Pony Ma stated that Tencent has consistently opposed using black-market plugins to record and transmit users' mobile phone and computer screens to the cloud, 'because it is extremely unsafe and irresponsible.'

After recognizing the numerous obstacles in the GUI simulation clicking technology path, Doubao proactively engaged in discussions with developers.

In January 2026, according to Smart Emergence, the Doubao Mobile team was in negotiations with mainstream app developers and had already secured some common permissions for internet products such as ride-hailing, food delivery, and ticket booking.

The newly launched consumer edition of Doubao Mobile Assistant has integrated services like Cao Cao Mobility and maps, enabling 'one-phrase direct access' through standardized service interfaces.

However, not all developers are keen on opening API interfaces to external AI tools. A case in point is Honor, which had previously focused on promoting protocol adaptations like MCP and A2A but encountered sluggish ecosystem expansion due to low developer enthusiasm.

If Doubao Mobile Assistant's consumer edition merely waits passively for official data interface adaptations, it may fade into obscurity.

02 Doubao Mobile Phone Still Utilizes GUI but Hands Control to Developers

It is widely acknowledged that since 2026, the focus of competition in the AI industry has shifted from large models to Agents. Compared to Chatbots that can only offer Q&A services, Agents capable of executing tasks hold greater potential.

The same principle applies to the mobile phone industry. In addition to enhancing capabilities like voice wake-up, memory, and screen recognition, the consumer edition of Doubao Mobile Assistant has not abandoned the selling points of Agents and task execution.

However, as mentioned earlier, with Honor serving as a cautionary tale, relying solely on a limited number of developers to provide API interfaces makes it challenging for the consumer edition of Doubao Mobile Assistant to achieve true 'one-phrase operation of the mobile phone.'

In light of this, the consumer edition of Doubao Mobile Assistant has adopted a 'dual-pronged strategy,' executing corresponding tasks through official data interfaces on one hand and relying on system-level GUI Agents as a backup on the other.

Image Source: OpenAI

In fact, bolstering GUI operation capabilities is emerging as one of the key technological directions in the Agent industry. For instance, to enhance the Agent experience, OpenAI has significantly boosted the GUI operation capabilities of GPT-6 Astra, enabling not only PCB design using KiCad but also house modeling in Blender.

The rationale is that after decades of development in the tech industry, there is a plethora of software applications, and not all developers are motivated or able to open interfaces. For long-tail software without open standard interfaces, GUI serves as a crucial fallback solution to expand the operational boundaries of Agents.

However, unlike computer software, which predominantly has tool attributes and involves fewer information security concerns, many mobile apps are service-oriented products. Most mobile internet companies are averse to GUI simulation clicking technology.

Image Source: Doubao

To avoid being blocked by third-party apps and sparking public controversy, Doubao Mobile Assistant simultaneously introduced the Screen Automation Execution Protocol (SAEP) and set a 30-day public notification period.

During the notification period, developers can independently declare whether to allow Doubao Mobile Assistant to operate their apps through GUI Agents. After the notification period expires, if developers have not informed via email or declined through the SAEP protocol, Doubao Mobile Assistant will enable GUI Agent operation capabilities for users.

Subsequently, if developers wish to adjust their app's interaction strategy with Agents, they can communicate with the Doubao Mobile technical team at any time.

It is evident that compared to the technical preview version, the consumer edition of Doubao Mobile Assistant has not made significant adjustments to GUI simulation clicking technology. The key change is that Doubao has ceded control over permission openness to developers. Compared to waiting for developers to open interfaces one by one, SAEP significantly reduces the coordination costs for GUI Agents to operate third-party apps.

03 Doubao Mobile Phone Commercialization: Starting with AI Office?

While Doubao Mobile Assistant's move to introduce the Screen Automation Execution Protocol is commendable, whether the product can instantly usher in the AI smartphone era is worth pondering.

Since 2026, the focus of Agent competition has primarily centered on office scenarios because users have clear cost-reduction and efficiency-enhancement needs in this context, and the benefits of Agents can be clearly quantified.

Unlike PCs, which possess strong productivity attributes, smartphones mainly cater to entertainment, social, and lifestyle service needs. Currently, about 700 million people spend two hours a day on Douyin, 400 million on Kuaishou, and 200 million on Xiaohongshu. The vast content consumption duration indicates that in the mobile internet context, users prioritize the 'emotional value' brought by content over faster operation efficiency.

In response, in November 2019, Alibaba's then-Chairman Jack Ma stated, 'About 17 million people browse Tmall and Taobao every night but buy nothing. I don't know what these people are doing; they're just aimlessly browsing. So your website can't just be for shopping; experience, fun, communication, and sharing are more important.'

The entertainment-oriented preferences of mobile internet users pose two major challenges for Doubao Mobile Assistant: first, the practicality of Agent functions is significantly diminished; second, commercialization becomes more arduous.

GUI Agents necessitate continuous screen recognition, operation path planning, and re-evaluation of interface states after each step, resulting in longer reasoning chains. Consequently, they consume more computational power than simply calling standard interfaces. In this context, since users do not have an urgent need for efficiency improvement, it is difficult for GUI Agents to establish a closed-loop commercial model by charging users.

Image Source: Doubao

Perhaps recognizing the structural contradictions in GUI Agent technology, the consumer edition of Doubao Mobile Assistant has emphasized bolstering office capabilities, supporting functions like recording minutes and work tasks. Users can command Doubao Mobile Assistant via natural language to perform tasks such as information retrieval, data analysis, and content generation.

Image Source: Doubao

In fact, charging users for office Agents is precisely Doubao's current focus. In June 2026, Doubao launched a professional version with standard, enhanced, and premium subscription plans priced at 68 yuan, 200 yuan, and 500 yuan per month, respectively.

From this perspective, what truly warrants attention in the consumer edition of Doubao Mobile Assistant is not just enabling Agents to further 'operate mobile phones' but also simultaneously touching upon two crucial boundaries for AI smartphones: one is the capability boundary determined by GUI Agents; the other is the commercial boundary based on office scenarios.

Over the past few years, AI smartphones have failed to become disruptive products not just because model capabilities were insufficient. A deeper issue is that for AI to complete tasks across apps, it must break through the long-standing ecological barriers between apps. Moreover, when task execution requires continuous calls to large models, visual understanding, and multi-round reasoning, the high computational costs deter manufacturers.

This is precisely where the value of the consumer edition of Doubao Mobile Assistant lies. GUI Agents based on SAEP can address interface fragmentation in a more gentle manner, while office functions attempt to find the first group of users willing to pay for expensive Agent capabilities.

If both paths can succeed simultaneously, AI smartphones could transform from a technological competition centered on models and functions into a scalable business.

Interactive Topic

Have you experienced an AI smartphone? What do you think of Doubao Mobile Phone?

This article is original content from Shijiao Farsight and is prohibited from reproduction without authorization.