At VivaTech, Datasection unveiled OHMAI, a physical AI companion capable of coordinating multiple models through a single interaction. Behind this device lies a broader question: What becomes of artificial intelligence when it ceases to be a chat window and becomes a physical presence?
It fit in one hand. Neither spectacular, nor metallic, nor deliberately futuristic. Rather, it was a small object placed on stage, halfway between a domestic robot and an office companion. Datasection, a Japanese company specializing in AI infrastructure, did not present OHMAI as a gadget. Its founder, Nori Ishihara, preferred a different way of putting it: artificial intelligence has reached a point where its main obstacle is no longer its ability to reason, code, analyze, or produce, but its lack of presence.
The statement might seem exaggerated. Yet it captures something true about the current state of AI. For the past two years, the industry has centered its competition on models: faster, larger, more multimodal, and more capable. But users don’t think in terms of models. They think in terms of tasks. They want to plan a trip, reply to messages, prepare for a meeting, make up a story for their child, track their health habits, or negotiate a service. They don’t want to choose between different types of intelligence. They want the intelligence to choose for them.
That is precisely what OHMAI promises: a small device that provides access, through a single account, to multiple AI models, automatically selecting the one deemed most suitable for the request. OHMAI Lite is priced at $9.99, OHMAI Pro at $99.99, with a monthly subscription fee of $3.99 for the account. The price is not a minor detail. It’s part of the vision: to make personal artificial intelligence no longer a privilege reserved for experts, but an everyday convenience.
Chatbot Fatigue
The chatbot has been the primary entry point for generative AI. It has made artificial intelligence visible, accessible, and almost familiar. But it has also imposed its own language: opening a window, formulating a request, choosing a tool, copying a result, prompting again, comparing, and adjusting.
The user is still the integrator. It is the user who carries the answers from one model to another. It is the user who decides whether to use a text, image, search, planning, or coding tool. It is the user who learns—sometimes in spite of themselves—to think like a machine operator.
Datasection raises an important point here: the problem with AI is no longer just the model itself, but the threshold for its use. As long as intelligence remains confined to an interface that must be activated, it remains intermittent. It intervenes only when called upon. It does not yet provide ongoing support.
OHMAI is trying to shift this mindset. AI would no longer be just a tab you open, but a presence you allow. A form of ambient intelligence, capable of memorizing habits, understanding communication styles, reading context, and then acting within both the digital and physical environments.
The Quiet Return of the Object
There is something paradoxical about this development. After fifteen years of smartphone dominance, artificial intelligence seems to be bringing the question of hardware back into focus. Why a dedicated device, when mobile apps already seem to cover everything?
Datasection’s answer can be summed up in one word: trust. According to Nori Ishihara, OHMAI was designed around principles that AI itself helped prioritize: accessibility, comfort, collaboration, adoption, and sustainable integration. Not to impress, but to reassure. Not to shine, but to become an integral part of the experience.
This rhetoric deserves to be viewed with caution. The idea of an “AI-guided” object also falls into the realm of theatricality. But it reveals a real shift: the next battle over AI will not be solely about power, but about personal acceptability. An assistant that knows your schedule, your messages, your preferences, your movements, and your habits must first be tolerated within your mental space.
Design then becomes less a matter of aesthetics than of diplomacy. If an object is too high-tech, it causes concern. If it’s too childish, it undermines credibility. If it’s too prominent, it’s distracting. If it’s too discreet, it becomes useless. OHMAI seeks that balance: a presence visible enough to be noticed, yet subtle enough not to impose its technology.
The Agent as an Operational Double
The most interesting part of the presentation wasn’t the final musical demonstration, but seeing agents talking to other agents. Scott Trowbridge, Chief Business Officer at Datasection, outlined a three-step path: standalone models, multimodal workflows, and then multisensory personal AI.
In this third phase, the agent no longer simply provides a response. He or she takes action. They sort through emails, prepare for a meeting, write in a tone that matches the user’s, coordinate schedules, negotiate with service providers, plan a trip, read a bedtime story, and adapt a school assignment to a child’s interests.
The promise is appealing: to reduce the invisible friction of everyday life. But it also raises a more challenging question. If the agent acts while the user is asleep, on the basis of whose preferences is it acting? How much leeway does it have? How much of the decision-making remains human? And above all, who controls the memory that makes this agent truly personal?
The luxury sector is already familiar with this issue, albeit in a different form. A great concierge is not measured by the number of services they offer, but by their ability to understand exactly what is needed without having to ask twice. They know when to step in, when to step back, when to offer assistance, and when to remain silent. Personal AI aims precisely at this fine line: transforming automation into attentiveness.
Details
OHMAI is based on a multi-model orchestration approach: the user does not choose which AI model to use. The system automatically selects the intelligence deemed most relevant for the requested task, through a single account priced at $3.99 per month. This is less of a model innovation than an interface innovation: the technical complexity is hidden behind a seamless user experience.
The New Interface War
Since the public launch of large language models, AI has often been viewed as a space for production: producing text, images, code, summaries, and ideas. OHMAI proposes something else: AI as a system that mediates between the individual and the world.
If this vision comes to pass, the question will no longer be simply which form of intelligence responds best, but which form of intelligence best represents us. The personal agent would then become a sort of secretary, butler, auxiliary memory, and discreet negotiator—an extension of the self, but one operated by a private infrastructure.
The phrase “AI for everyone” is appealing. However, it comes with a catch: truly personal AI cannot be merely practical. It must be governable. Users must be able to know what is retained, what is discarded, what is shared, what is decided independently, and what requires validation.
So the next frontier in artificial intelligence may not be the most powerful model. It will be the most trustworthy model—the one that stops merely responding within a chat window and starts taking action in our lives.
OHMAI may be just a first step. But it’s already asking the right question: when AI finally steps off the screen, will we still be able to distinguish between assistance and delegation?
Cette publication est également disponible en :
