The mobile the AI ​​promises, but I only see repeated tricks. The real ace under the sleeve is called “agent” and comes on the way

How could it be otherwise, AI has also reached our mobilemainly from the hand of two protagonists: Apple Intelligence on one side, Google Gemini In the other. However, in a matter of one year, we have gone from almost no manufacturer has ia that many protagonists of the industry now fight in the classic race.

Samsung was a pioneer in some Galaxy AI’s first steps in which it enveloped the generative AI with profits: transcription voice recordings, simultaneous translation and generations of wallpapers, to name a few. More than a year later, AI suites for smartphone have multiplied. And yet, There is still a long way to go.

Although these tools are a first step, the true pending revolution is the agricultural AI, capable of performing complex tasks autonomously. However, its implementation faces significant obstacles that limit their current potential.

The next step is the agents

After witnessing the MWC 2025, it is clear that IA has not been the central theme, although the most mentioned. Obviously, he has had his time of prominence, with Rimbombantes “AI” on several fronts: Motorola, Honor, Realme and Samsung have presumed suites for Android.

Magicai, Next AI, Hyperai … prolong the journey of Galaxy AI, which by the way, has also overcome herself with the arrival of the Galaxy S25. How is better about the first -time version? We basically talk about two more skills, which revolve around to personalization. A short but sure step, that in essence it does not be far from What I could test Last year.

Of those mentioned, the one that surprised me the most was Magicaithe honor suite (integrated into magicos) and its ambitious Honor Alpha Plan. For a simple reason: already introduces the concept of AI agents.

It is not only an AI that responds, but one that can understand a complex objective, plan steps and execute actions proactively through different applications and services to achieve it (for example, plan and reserve a complete trip, not only look for a flight).

These agents or agriculture contrast with current functions, which are more reactive or limited to a single task or app. Google has taken some shy steps with Gemini’s extensions, But there is a problem to treat: Interoperability between services.

For an agent to reserve a table or ask for food, he needs to interact fluently with any restaurant or cast app, which in turn requires standardized APIS or a deep level of integration that, for the moment, does not exist in a generalized way. On the other hand, these demonstrations have an important limitation: they control the screen and do not allow the user to continue using the mobile while the agent acts. Leave that feeling that the machine does it for us, but it blocks us the machine itself.

In the demonstration, the Honor’s AI could interact with the phone and the real world: Given the information that the brightness of his screen was very high, the mobile responded with the instructions while reducing it. Similarly, he is able to book a table in a restaurant, running its respective app automatically.

Surely that is the correct direction of the mobile AIalthough it is early to say it. Along the way, several actors must be taken by the hand: not in vain, asking for online food is not just Google or Honor, in this case. Now, not only these agents are useful, but also the rest of the tools that manufacturers are adding to their customization layers.

Galaxy ai is not the only

It is easy to highlight the aforementioned Samsung Galaxy AI: it was a pioneer to incorporate tools based on generative, yes, together with Google. ‘Roda to search’ opened the way and is now present in a large number of Android mobiles. That is why both the Pixel and Galaxy, and the rest of the manufacturers, now have some Common functions. That are also far from the aforementioned agents.

The difference marks the customization layer. Out of what Google lends its partners, the interesting thing is to know the commitment of manufacturers such as Honor, Xiaomi, Realme who already tries to follow Galaxy AI’s track. This is what I found.

I start with the most unknown in Europe, a techno that could perfectly rival the strongest in the sector. Your techno aidoes not move away from what is seen in other suites of tools: restoration of old photographs, transformation of portraits into animations, a call assistant. In short, nothing new under the sun.

Xiaomi, who has announced the arrival of his Hyperai In the new ones Xiaomi 15also proposes a similar conjunction: Ai Writing generates messages for us, Ai Speech Turn the voice into text, it has a translator in real time, generation of images, draft objects (example photos) and even a function destined to the recording of film video.

Is more extensive than in other manufacturersbecause Xiaomi already had this AI suite in China for a long time. As we see, photography is one of the most handled aspects by AI, along with productivity tools. There are others that try to differentiate themselves, among them the mythical Motorola, now under the amparo of Lenovo.

Moto AI It is the name that receives its AI system, and not only affects the phones: Motorola adds its own agent to the most advanced headphones of the catalog. I witnessed a demonstration that pointed ways, no matter how much had some timely delay, given by the Internet connection.

This system includes a voice assistant in the purest Gemini style: capable of organizing the agenda, saving annotations in a personal diary and of course, improvements in the photographic processing to avoid blurred shots. For now, Motorola does not have many models compatible with motorcycle AI: only the Edge 50 Ultra and the Razr 50 and Razr 50 Ultra.

Realme, and his NextaiThey look similar to the AI ​​application to the mobile phone. Hyperimage+ It is part of it, as a set of algorithms capable of improving photography. Like others, add tools to eliminate unwanted people, summarize voice recordings and transform sketches, in addition to a curious Ai smart loop which allows to drag any element on the screen to take it to other applications (for example to share an image on WhatsApp, from the browser).

As you have seen, most of these functions are common, often derived from Google capabilities or replicas of other existing ones. Despite their concrete profits, they do not represent what is expected in the near future: they are collections of functions, not proactive or autonomous systems. No manufacturer, neither does Samsung, have implemented a truly large -scale agriculturaldue to the limitations I mentioned.

I feel that (and must) give much more of itself

After knowing in depth all the software battered with the proposed by the various signatures, a clear conclusion about the current state and its future potential emerges. While functions such as instant translation, transcription or photographic edition are useful in specific situations – such as being abroad or improving an image – it is easy to perceive that most suites offer similar capabilities.

However, repeating me, the true qualitative leap that would transform the mobile experience resides in the agricultural AI: systems capable of understanding complex user objectives and acting proactively and autonomously through different applications and services to achieve them. The reason why this agricultural vision is not yet a generalized reality, and by which many current suites may feel limited or iterative, lies in fundamental technical and ecosystem obstacles:

• Limited interoperability: Applications do not easily communicate with each other, making it difficult for agents to perform complex tasks that involve several services.
• Execution in the foreground: The agents seen control the screen and block the use of the mobile, instead of operating discreetly in the background, which requires greater integration with the operating system.
• Cloud and power dependence: They require constant Internet connection for many tasks. The local AI is in diapers compared.
• Privacy and safety: Giving autonomous access to data and apps to an AI will be the next concern for the user. It entails important risks that need controls and transparency.

That is still unknown when it will be useful for the bulk of usersoutside of concrete situations such as being abroad or the subtle pleasure of turning a scribble into a beautiful image. The AI ​​for the mobile needs to take another step, to be useful for each moment: in this sense, I think that Samsung has given a step, but yes, one rather short With Now Brief.

Of the rest, yes, for situations in which productivity prevails, they are useful. However, nothing new that we do not know about the generative AI, and that Chatgpt itself cannot do. Then, there are the promises around AI for photography, little tangible, except for some like the tremendous zoom of which OPPO presumes and also honor. Yes indeed, They blur with the image process of the Android cameras themselves.

I am not and I will be AI denialist, much less on my favorite device. However, for mobile AI to deploy its potential possibilities and become a truly transformative tool in daily use, it is necessary to address these fundamental limitations. Perhaps in the next generation we will see more significant advances towards an ecosystem that allows the agents to operate in an integrated, autonomous and useful way at each moment, fulfilling the promise that only is glimpsed today. The current state, although promising on some fronts, It is far from that agricultural future.

In Xataka | Chatgpt as a digital wizard on Android: how to configure it to replace Gemini

In Xataka | How to avoid responses generated by Google with this simple trick