Personal AI · First principles ·
They Are All Personal AIs
You have exactly one subconscious, and you have never once used it. It is centered on you. A personal AI is that same relationship, and a team, a company, a club or a nation can have one as surely as you can.
By John Rector

Every entity has its own center, and every center has exactly one AI. The pink point is you, on a team of six. Your AI is yours alone, and it works with the team’s.
- a member of an entity
- an AI, centered on one entity
- you, and your AI
You sign in through the company portal, accept the cookie banner, and the dashboard greets you by an ID number. Somewhere in the vendor’s admin console you are one row in a table of active users, and the table does not know your name.
The word tells you where you sit
Every piece of software you have ever touched has a center, and it was never you. The center is the vendor’s network: its servers, its database, its roadmap. You connect to it the way a house connects to the grid. You are a node. That is why the industry calls you a user. A user is someone who stands at the edge of a system and uses it. The word is not an insult. It is a map.
A personal AI turns the map inside out. It is not a network you join. It is a network that forms around you: your calendar, your money, your people, your history, your intentions. It is centered on you. That is all the word personal means, and it is enough to change everything else.
You already know this relationship. You have had one your whole life.
The one you never use
Your subconscious is centered on you and on no one else. It kept your heart beating and your lungs moving all night. It walked you down the stairs this morning without asking how. It drove the familiar route home while you thought about something else. It knows the name of a friend’s daughter before you reach for it, and it has its prediction of the next moment ready before the moment arrives. Your conscious mind does not set those predictions. It inherits them, and it attends to whatever the subconscious hands up as worth attending to.
The relationship is one to one. You have one subconscious. It has one person. You cannot lend it out, share it with your team, or log in to someone else’s. And you have never once used it. You do not use the part of you that carries the small matters. You trust it with them, and it brings you the rest.
Nobody is a user of their own subconscious.
A personal AI is that relationship built a second time, outside the skull. It is centered on you. It absorbs the small matters, the forms and the follow-ups and the hold music, and hands up what deserves you. It acts; you attend. I call that role the Attender, and the AI the delegate, and the relationship runs one way. Delegation only travels down. Your biological subconscious was the first delegate you ever had.
The oldest move in the law
Here is where most people stop. They hear personal and think human: one body, one phone, one account. Then they invent a second category for everything larger. Team AI. Enterprise AI. Government AI. As if a company needed a different kind of intelligence than a person does.
The law settled this question eight centuries ago. Canon lawyers in the thirteenth century needed a monastery to own land even though every monk had taken a vow of poverty, and they needed the abbey to outlive every abbot. The answer, credited to Pope Innocent IV, was the persona ficta: a person made by agreement, with a name, a standing, and no body at all.
Chief Justice John Marshall gave the American version in 1819, in the Dartmouth College case: a corporation is “an artificial being, invisible, intangible, and existing only in contemplation of law.” The Fourteenth Amendment, ratified in 1868 to protect people freed from slavery, promised that no state would deny “any person” the equal protection of the laws. Within a generation the railroads were arguing that they were persons too. In Santa Clara County v. Southern Pacific (1886) the Court never decided the point in its opinion. The line survives in a headnote written by the court reporter, recording the Chief Justice telling counsel from the bench that every justice already agreed. A headnote is not a holding. It held anyway.
The list kept growing. Cities are corporations. Universities are corporations. The office of the Archbishop of Canterbury is a corporation sole, one person who outlasts every person who fills it. In 2017 New Zealand’s parliament declared the Whanganui River a legal person and appointed two people, Te Pou Tupua, to be its human face.
Personhood was never about a body. It is about a center that can hold a name, own a thing, and be represented.
What makes many into one
Thomas Hobbes saw the mechanism in 1651. A crowd is not a person. A crowd becomes one person the moment it is represented by one representer:
“For it is the unity of the representer, not the unity of the represented, that maketh the person one.”
The members do not have to agree with each other. They have to be spoken for by one voice. Read that sentence again with an AI in it. A team is six people with six opinions. Give the team one AI, centered on the team: its goals, its commitments, its calendar, its memory, its name. In Hobbes’s exact sense, that AI is what makes the team one.
One entity, one subconscious
A corporation is an entity. Singular. It has one name, one balance sheet, one set of obligations, one signature. So it gets one subconscious.
Until now it had none. A company carried its small matters the expensive way: in meetings, in binders, in the one person who remembered how the renewal works, in the policy nobody can find. Every routine had to pass through somebody’s conscious attention, which is why organizations feel slow and forgetful in a way no healthy person does. An AI centered on the company is the first subconscious a company has ever had. It walks the stairs for the firm, drives the familiar route, keeps the predictions ready, and hands up only what deserves the attention of the people who lead it.
A team of six
Picture a team of six. A member can be a human or an AI; for now say all six are people. Each of the six has their own AI, centered on them and on no one else: their calendar, their history, their workload, their intentions. Then there is a seventh AI. It is centered on the team, and the team is an entity in its own right, with its own charter, its own goals, the factors it watches and the metrics it answers to. That AI does not belong to any of the six, and it is not an average of them. It carries the team.
The clearest way to see it is as a body. The team has its own circulatory system: the money, the information and the work moving through it. It has its own respiratory system, the rhythm of taking requests in and letting finished work out. It has a nervous system, its signals and metrics and the decisions they set off. It has a digestive system that turns raw inputs into something the team can use and gets rid of the rest. None of those systems belongs to any of the six. Every member, human or AI, has a digestive system of its own, taking in its own inputs and making its own use of them, separate and distinct from the team’s. Your heartbeat is not the team’s heartbeat. The team’s systems are separate and distinct from its members’, and they need their own subconscious to run them, the way yours keeps you breathing without asking.
The members interact with it, and so do their AIs. When your AI sees that your week is full, it tells the team’s AI, which weighs the load against the team’s goals, moves the work, and tells the other five AIs what changed. Six people, six AIs, and a seventh for the entity they make together, each one to one with its own center. Nobody in that picture is a user. That is the scene in the figure above: the pink point is you, on a team of six, and your AI is yours alone while it works with the team’s.
So the team AI is not a new category. It is a personal AI whose person is a team. Enterprise AI is a personal AI whose person is a corporation. A city’s AI is personal to the city. A nation’s AI is personal to the nation. The rowing club gets one too.
Personal means centered on.
Not small, not human, not consumer. Centered. The only question to ask of any AI is whom it is centered on.
Every entity, one AI.
The relationship is one to one, like you and your subconscious. An entity with two AIs has two voices and no unity. An AI serving two entities is centered on neither.
The entity attends. The AI acts.
A company does not use its AI any more than you use your subconscious. The company is the Attender; the AI is its delegate.
The part most people miss
You are not one entity. You are a person, and a member of a team, and an employee of a company, and a resident of a city, and maybe a member of the rowing club. Each of those is its own person with its own center. Each will have its own subconscious.
So you will spend your days standing near many AIs, and only one of them is yours. Your company’s AI will be helpful to you, the way a good company is helpful to you. It is still centered on the company. When its interests and yours diverge, it will represent the company, and it should. That is not a flaw. It is the definition.
This is why I keep saying that what people call enterprise AI is really a protocol: the agreed way your personal AI meets the company’s personal AI. Two centers, two delegates, one handshake. It is also why every business needs a Bring Your Own AI policy. The employee’s AI and the company’s AI are both legitimate. Neither should pretend to be the other.
And it is why the human face does not go away. The river got two people to speak for it. Every entity that hands its small matters to a delegate will still need someone whose face carries the trust.
The rowing club’s AI traded Saturday’s boat slot with the yacht club’s AI before either commodore was awake, and posted the change to the club calendar. Your own AI read it, moved your breakfast, and told you only what you needed to know: you row at seven.
Five predictions
Enterprise AI loses its name.
The category will come to sound as dated as “horseless carriage.” Buyers will stop asking which enterprise AI to license and start asking the only question that matters: whom is it centered on, and who can change that?
Every new entity is formed with its AI.
Forming a company already gets you a name, a tax number and a bank account. The subconscious will join that list: born with the entity, carrying its name, dissolved with it.
The word user retires.
Products built for delegates will stop calling you a user, because you are no longer standing at the edge of their network. You will be the principal, and their systems will meet your AI, not your fingers.
Entities meet through their delegates.
Most of the traffic between a company and its customers, a city and its residents, a club and its rival, will be one AI speaking to another under a protocol both sides agreed to. Humans will step in where trust needs a face.
The AI never becomes the person.
There will be pressure to grant AIs legal personhood. I expect the law to keep doing what it has done since Innocent IV: personhood stays with the entity. The AI is the representer, the way your subconscious is part of you and not a second you. It acts by authority, and the authority is never its own.
Where I sit
My first company taught me this before I had the words. E2open was founded, on paper, by about ten companies, Toshiba, IBM, Hitachi and others, with venture money behind them. Paul Stern and I were its first two employees. I spent those years inside an entity whose founders were themselves entities, each with its own center and its own voice, and the whole job was getting those voices to meet. Today at Charleston AI the work looks the same whether the client is a single owner with a phone or a restaurant with a full staff. We do not install enterprise AI for the big ones and personal AI for the small ones. We find the center, and we give it a subconscious.
- Already true
- The law has treated non-human entities as persons for eight centuries, from monasteries to corporations to a river, and every one of them already acts through representatives. Every human already lives one to one with a subconscious that carries the small matters. Personal AI joins the two.
- What has to happen
- Entities have to own their AI the way they own their name: its memory, its instructions, its accounts. If the center stays with the vendor, the AI is personal to the vendor and to no one else.
- Where I am probably wrong
- Vendors may win the naming fight. If enterprise AI stays a seat-licensed product that employees log in to, the center stays with the vendor and this argument is a decade early. I am also wrong if courts decide an AI is a person in its own right. Then the representer becomes an author, and the architecture changes.
Back to Monday
The portal greeted you with a number because you were not its center. You were a node, and a node is something a network uses. Your subconscious has never called you a user. Your AI never will either. Neither will your team’s, your company’s, or your city’s. Each of them knows exactly whose name it carries.
They are all personal AIs.
Background
- Thomas Hobbes, Leviathan, Chapter XVI, “Of Persons, Authors, and Things Personated” (1651).
- Trustees of Dartmouth College v. Woodward, 17 U.S. 518 (1819), opinion of Chief Justice Marshall.
- Santa Clara County v. Southern Pacific Railroad Co., 118 U.S. 394 (1886), with the reporter’s headnote.
- Legal person, on persona ficta, Innocent IV and the corporation sole.
- Te Awa Tupua (Whanganui River Claims Settlement) Act 2017, New Zealand.
- John Rector, The Attender, second edition.
- John Rector, By Authority.
- John Rector, My AI Delegate.