What Happens When Individuals Can Suddenly Think Institutionally?
Institutions have a structural advantage.
That is neither surprising nor inherently objectionable. A public authority, an insurance company, a bank, a hospital, or a corporation has access to expertise, professional language, records, procedures, defined responsibilities, legal advice, and above all: paid working time.
On the other side stands an individual.
Often precisely when that individual is not doing particularly well. When they are ill. When they have lost their job. When money is running out. When an application has been rejected. When they are confronted with a diagnosis they do not understand. When they are expected to sign a contract whose consequences they can barely assess.
Our societies are aware of this asymmetry.
A significant part of our legal systems can even be understood as an attempt to limit it.
Citizens can challenge administrative decisions. Patients are supposed to give informed consent. Consumers have rights to information and withdrawal. People can request information, inspect records, have decisions reviewed, and take matters to court.
It sounds like a relationship between equals.
In practice, it often never was.
The Strange Ideal of the Empowered Individual
Many of these rights presuppose an astonishingly capable human being.
They would first have to understand the letter.
They would have to recognize which statement in it actually matters.
They would have to know what information is missing, distinguish facts from assessments, find previous correspondence, identify contradictions, observe deadlines, and then formulate their objections in a way the institution can process.
When the answer arrives, they would once again have to recognize whether their actual question has been answered.
And all of this alongside their ordinary life.
Someone who cannot do this has not lost their rights. But they may have lost a substantial part of their ability to actually exercise them.
In many places, the law therefore postulates an empowered, informed, and capable individual who, in reality, exists only under favorable circumstances.
Or with professional help.
Lawyers, tax advisers, doctors, social associations, trade unions, consumer organizations, and many others therefore perform an important function: they translate between individuals and institutions.
Until now, that translation has been scarce, expensive, or at least associated with considerable effort.
AI changes precisely that.
Suddenly, Someone No Longer Loses the Thread
Perhaps the most socially interesting property of a large language model is not that it can write texts.
Nor is it that it “knows” a great deal.
It is something far less spectacular:
With its help, an individual can begin to think institutionally.
They can dissect a ten-page administrative decision.
They can compare a medical report with an earlier one.
They can ask whether an answer actually answers their question.
They can have a technical term explained without having to become an expert themselves.
They can write an angry email — and not send it.
Instead, they can spend two hours discussing with an AI what is actually making them angry, which parts of their interpretation are supported by the available information, and which are not.
In the end, they may send three polite paragraphs.
Not because the AI made the decision for them.
But because it helped them determine their own position more precisely.
That is a fundamental difference.
Rights Suddenly Become Liquid
Perhaps we have underestimated the practical consequences of this.
A right does not derive its value merely from being written into law. Its holder must also be able to mobilize it.
That costs knowledge, time, attention, money, and psychological energy.
AI dramatically lowers some of these transaction costs.
And something peculiar can happen as a result:
Rights that have existed for a long time suddenly alter the actual balance of power.
Not because legislators have created new rights.
But because many more people can understand, articulate, and persistently exercise the rights they already have.
“You are, of course, entitled to appeal” means something different when the recipient then actually submits a precise and well-reasoned appeal.
And when the response is incomplete, an equally precise follow-up arrives.
And when that response differs from an earlier one, the earlier one is still there.
The individual acquires something that until now was primarily the preserve of organizations:
memory, structure, and persistence.
What If Everyone Does This?
This is where the problem for institutions begins.
Many administrative and decision-making processes function partly because the overwhelming majority of people do not fully exercise all the possibilities theoretically available to them.
That is not necessarily intentional.
It is simply a reality of scale.
An institution can handle a hundred imprecise inquiries through standardized procedures.
But what happens when those same hundred people suddenly identify the decisive question?
When they read the answer?
When they notice that their question was not answered?
When they follow up?
When they compare justifications?
When they do not give up, exhausted, after the third letter?
Then real work is created on the other side.
And, remarkably, we currently find ourselves in a transitional period in which this problem is particularly pronounced.
Institutions Often Cannot Do What Individuals Already Can
The obvious answer seems straightforward:
Then both sides will simply use AI.
In the long run, they probably will.
Today, however, the situation is often very different.
A citizen can sit at home and use ChatGPT to analyze a medical report, an administrative decision, or their own correspondence.
The employee inside the institution may be prohibited from entering the very same case into a public AI system.
For good reasons.
Data protection. Professional confidentiality. Trade secrets. Information security. Compliance. Documentation requirements. Auditability. Liability.
An institution cannot simply say:
Let’s try it.
It needs governance.
And so, at least temporarily, a remarkable reversal emerges.
For decades, the institution possessed the superior information apparatus.
Now it may find itself facing an individual who, with an AI, can spend days working exclusively on one single case.
The caseworker cannot do that.
The doctor cannot do that.
The insurance employee cannot do that.
Not because they are less competent.
But because this one case is one of many for them.
For the individual, it is their life.
This Is Not Automatic Democratization
None of this should be turned into a romantic story.
AI does not automatically make people wiser.
It can structure errors just as brilliantly as legitimate objections. It can help someone defend a false belief with increasingly sophisticated arguments. It can create false confidence. It can flood institutional processes with perfectly formulated nonsense.
And a person arguing with the assistance of AI is obviously not therefore right.
That would merely create the next dangerous asymmetry.
The interesting change lies elsewhere:
They can suddenly sustain a reasoned position.
And they can subject it to scrutiny.
A good AI does not merely generate arguments on their behalf. It can ask:
How do you know that?
What other explanation might there be?
What does the document actually say — and what are you reading into it?
What information would change your assessment?
Used in this way, AI does not replace autonomy.
It may make autonomy practically possible in the first place.
The Next Escalation Is Easy to Predict
Institutions will, of course, catch up.
Then the citizen’s AI will analyze the administrative decision while the authority’s AI analyzes the appeal.
One AI drafts the follow-up.
Another AI drafts the response.
Technically, this can scale beautifully.
Socially, it could become absurd.
Because the decisive question would no longer be who can argue faster or more eloquently.
It would become:
Does any of this actually clarify the matter?
Or are we merely automating, on both sides, the ability to defend positions against one another with ever greater perfection?
That would not be the end of bureaucracy.
It would be its industrialization.
Perhaps Institutions Need to Learn Something Else
The long-term answer can hardly be to make individuals weaker again.
We will not realistically be able to prohibit people from using AI to understand a medical report, a contract, or an administrative decision.
And we probably should not want to.
Instead, institutions will have to be designed for a world in which the person on the other side may actually be as capable as our legal systems have always pretended they were.
A patient who asks questions.
A citizen who reads the reasoning.
A consumer who compares contractual terms.
An applicant who recognizes a contradiction.
A person who remembers what they were told in writing six months ago.
None of this should constitute an attack on an institution.
It is the realization of what we have been calling individual agency for decades.
The Real Shift
When we talk about AI, we spend a great deal of time discussing which human activities machines will take over.
Perhaps we are overlooking another transformation.
AI can do more than make institutions more powerful.
It can give individuals capabilities that previously had an almost institutional character: memory, research, structure, linguistic precision, translation across domains, and the persistence required to follow a complex matter consistently over long periods of time.
That shifts power.
Not necessarily toward truth.
Not automatically toward the individual.
And certainly not without creating new problems.
But the old asymmetry can no longer be taken for granted.
Perhaps one of AI’s most consequential effects on society will therefore not be that machines eventually learn to act autonomously.
But that humans suddenly become better able to do so.
What happens when individuals can suddenly think institutionally?
We are only beginning to find out.
Comments
Write a comment