Imagine someone saying today,
“I don’t have email” or
“I don’t have a phone.”
Neither is impossible, but both sound unusual because these have become basic parts of modern life.
I believe “I don’t have AI” will eventually sound much the same.
At the moment, that sentence can still simply mean that someone does not actively use ChatGPT, Claude, Gemini or another AI service. But AI is rapidly becoming less like a separate category of software and more like a basic capability that becomes embedded into how people think, write, research, learn, analyze, create and work.
The more that happens, the less interesting the question becomes whether someone uses AI at all.
The more important questions are what kind of AI they use, what it knows about them, what it can do for them and, eventually, whether that AI is actually theirs.
That is where the discussion starts to shift from AI in general to personal AI.
There is a fundamental difference between having access to AI and having your own AI.
A free AI service is not really yours.
An AI provided by your employer is not your personal AI either.
Both may be extremely useful, but they exist inside someone else’s environment and under someone else’s rules.
A personal AI is different because it has an ongoing relationship with you.
It knows the information you choose to give it, understands your preferences, develops context about your work and your life, and can become more useful over time because it increasingly understands your world. As these systems become more capable of taking action on our behalf, the idea also moves beyond a simple assistant and starts to become something closer to a personal AI agent.
For me, the best analogy for this is not the personal computer. It is the mobile phone.
Some adults can still reasonably say that they do not need a personal laptop. Very few would say the same about having their own phone.
Now imagine if your phone worked in a way where you could never change the hardware manufacturer or mobile operator, and changing provider meant losing your number, contacts, messages, applications and history. That would sound absurd.
The mobile ecosystem has evolved around the opposite idea.
Devices can change, operators can change, applications can change and the underlying technologies can change, but your identity remains connected to you.
Your phone number can follow you and your information can move with you. The surrounding providers are replaceable.
I believe personal AI will ultimately need to follow the same principle.
Models will change, providers will change, agent software will change, applications will change and devices will change, but your identity and the information associated with you should remain yours.
This is where the question of information becomes central, because when you look closely at what actually makes a personal AI personal, almost all of it comes down to information.
The name you give it is information. The role you define is information. The behavior you expect is information. Its instructions, memory, goals, permissions and access rights are all information.
The people, organizations and things it knows about are information. Even the skills it has are, in many cases, simply information expressed as instructions, workflows, examples, rules or procedures.
The software and models are obviously essential, but without the right information they do not know who they are working for, what matters, what they are allowed to do or how they should behave.
Software, including the AI agents themselves, is replaceable. New versions will come and go. But even a highly capable AI without access to your information is clueless about you and your world.
Information is what turns generic AI into your AI.
This is also why ownership and portability become so important. Today, discussions about digital ownership usually focus on files, photos, messages, health information, financial information and other forms of data. With AI, the scope becomes much broader because we also start accumulating information about how our AI works with us.
Over time, you may refine how it communicates with you, teach it your preferences, build workflows around it, define rules, add specialist skills, connect it to services and give it different access rights. You may gradually build years of useful context around your personal AI.
If all of that disappears when you change providers, then the system is not really portable in any meaningful sense.
Imagine using your personal AI for five years and then deciding that another provider has a better model. If moving means starting again from zero with no memory, no preferences, no skills, no relationships, no permissions and no understanding of your work, it would increasingly feel like changing mobile operator and being told that your phone number, contacts and messages cannot come with you.
In future that decision will be harder. Today it is still easy.
The model should be replaceable. The provider should be replaceable. The application should be replaceable. The information that makes the AI yours should be separate.
Once personal AI is viewed this way, several larger questions appear.
- How do we maintain a persistent digital identity?
- How do we retain ownership and portability of our personal information?
- How do we move the accumulated information that makes our AI useful from one provider to another?
- And how do we define the boundaries between our AI and the AIs of other people, companies and services?
That last question may become especially important because personal AI will not exist in isolation.
Your AI will interact with other people’s AIs, your employer’s AI, your company’s AI, the AIs of services you use, community AIs, banks, healthcare providers, government systems and many other organizations. A growing share of digital interaction may eventually happen between agents acting on behalf of people and organizations. In that world, identity, permissions, trust, ownership and boundaries become foundational.
One possible part of the answer already exists in a very familiar technology: email.
An email address is still one of the closest things we have to a universal digital identity. It sits underneath much of our online life and, importantly, it can actually belong to you. If you own your domain, you can change email provider without changing your identity.
That provides a useful model for how personal AI could work. Your identity should be portable and your information should be portable.
The personal information core around your AI should ideally be based on open standards, with open-source implementations available, so that it can be hosted in different ways without losing its connection to you.
You might host it yourself. A telecom operator might provide it. A bank might provide it. Your employer could provide a separate work-related environment. A specialized AI operator could provide it. Or you might simply buy it as a managed service.
The important thing is that the provider should not become the owner of your digital self. Changing provider should not mean losing who you are or everything your AI knows about you.
The more I think about this, the more AI adoption starts to resemble earlier technology transitions. General AI skills look a lot like the early internet and email.
At first they are used mainly by specialists and curious people, then they become useful work skills, and eventually not knowing how to use them starts becoming a disadvantage.
Personal AI looks more like the mobile phone transition. In the early days, personal mobile phones were disproportionately used by entrepreneurs, executives and businesspeople. Then adoption broadened. Eventually the question was no longer whether people should have one, but how early children should get their first phone.
Expect something similar to happen with personal AI.
People who can build their own are already doing it. Others are assembling them from open-source components, while others use increasingly personalized commercial services.
Many people may not yet describe what they want as a personal AI, but the underlying demand is easy to recognize.
People want AI that remembers, understands context, knows their preferences, works across services and can increasingly act on their behalf.
Once those expectations become normal, the distinction between “an AI I can access” and “my own AI” becomes much more important.
That is also when an entirely new infrastructure market starts to appear around personal and organizational AI.
There will be providers, operators, identity systems, portability requirements, interoperability mechanisms, security layers and regulatory frameworks.
I expect we will eventually see something resembling AI operators: companies that provide persistent AI infrastructure for people and organizations in much the same way that telecom operators, banks, email providers and cloud providers serve other critical parts of our digital lives.
For entrepreneurs and owner-managers, this creates a particularly interesting window.
Right now, you can still learn this personally before the structures around it become standardized. You can understand how AI changes your own work, how a personal AI becomes useful over time and what ownership actually means in practice.
From there, the same questions naturally extend to the company.
- What information belongs to the company?
- What belongs to employees? What context should the company AI retain?
- Which workflows and skills should remain with the company?
- What permissions should different agents have?
- How should the company AI interact with customers, suppliers and partners?
- What happens when an employee leaves, or when the company changes AI providers?
Today, relatively few organizations have clear answers to these questions. Later, they may simply become normal business architecture.
For professionals, the transition is just as important. Learning AI is not only about becoming faster at your current work. It is increasingly about understanding how your expertise can be captured, reused and scaled.
A useful process can become an AI skill. A way of researching something can become a reusable workflow. A decision framework can become a set of instructions. A professional method can be represented in a form that an AI can repeatedly apply.
Skills, in other words, increasingly become information too.
That does not make the software irrelevant. It simply highlights where much of the long-term value sits. The model provides intelligence, but information gives that intelligence direction, context and meaning. It defines what the AI is, what it knows, what it can do and what it is allowed to do.
That is why I think the future of personal AI is ultimately also a question of information ownership.
Nobody can do the learning part for you either. A company can give you tools, a school can provide courses, a coach can help and an AI can teach you, but someone else cannot acquire the skill on your behalf.
It is like going to the gym. A personal trainer can show you what to do, but they cannot grow your muscles for you. Someone can explain how to swim, but at some point you still have to enter the water.
AI is similar. Interestingly, the best tool available for learning how to work with AI is AI itself. Others can help you understand what matters, avoid mistakes and progress faster, but ultimately you have to use it.
I would therefore not treat personal AI as some distant science-fiction idea. Treat it as a hypothesis and then simply observe what is happening around you.
Watch what people start expecting from AI. Watch what developers are building. Watch how quickly AI systems gain memory, tools, agents, identity and access to personal information. Watch what happens when people become frustrated because their history, context and workflows are trapped inside one provider.
Then ask whether this direction is becoming more or less likely?
If you are an entrepreneur, investor or business leader and believe this direction is even approximately correct, there are major opportunities in the infrastructure around personal and organizational AI.
If you are a regulator, there is an equally important responsibility. The last few decades of digital development have concentrated enormous amounts of identity, information and control inside a small number of platforms. AI gives us an opportunity to reset and build the next layer differently.
Personal AI should not mean that one company owns an increasingly complete digital representation of you.
Your identity should remain yours, your information should remain yours, and everything that makes your AI truly yours should be portable.
Because the models and software will keep changing.
The information is what gives them meaning.
And if this develops the way expected, there may eventually be one sentence that sounds just as strange as saying you do not have email or a phone:
“I don’t have my own AI...”
Originally published on LinkedIn.