Part 4 of After Intelligence: From Capability to Agency in the Age of AI**
One of the most attractive promises surrounding artificial intelligence is democratisation.
Capabilities once restricted by expertise, money or institutional scale are becoming available to almost anyone with an internet connection.
You do not necessarily need a software team to prototype an application.
You do not need a design department to visualise an idea.
You do not need a research assistant to summarise a field.
You do not need a marketing agency to draft a campaign.
AI can dramatically lower barriers.
That is real.
It is important.
And it is worth celebrating.
But there is a dangerous step hidden inside the democratisation argument.
We can easily move from:
Everyone can access more capability
to:
Everyone will therefore benefit equally from that capability.
Those are very different claims.
The same technology placed in different hands does not necessarily produce the same possibilities.
And this raises what may be one of the most important social questions of the AI era:
What if AI democratises capability without democratising agency?
Same capability. Different agency.
Imagine two employees in different organisations.
Both receive access to exactly the same powerful AI assistant.
The first works in an environment where:
- experimentation is encouraged;
- managers welcome questions;
- mistakes are treated as learning;
- there is time to explore;
- colleagues share what they discover;
- good ideas can reach decision-makers; and
- employees are trusted to improve things.
They ask AI:
“What could we do differently?”
The second works in an environment where:
- tasks are tightly prescribed;
- performance is closely monitored;
- mistakes are punished;
- experimentation creates risk;
- workloads leave little time for exploration; and
- decisions remain highly centralised.
They ask AI:
“Can you help me finish this faster?”
Same technology.
Same nominal capability.
Very different agency.
The difference is not inside the model.
It is in the surrounding human system.
Access was only the first digital divide
We have seen versions of this before.
Early discussion of the digital divide often focused on physical access.
Who has a computer?
Who has broadband?
Who can connect?
Those questions were and remain important.
But as access expanded, deeper divides became visible.
Who has the skills to use technology effectively?
Who understands how information systems work?
Who can judge online information?
Who can turn digital access into educational, social or economic advantage?
AI is likely to follow a similar pattern.
The first divide is obvious:
Who has access to powerful AI?
The second is already emerging:
Who knows how to use it?
But the third may matter even more:
Who has the agency to convert it into meaningful action?
Agency has prerequisites
We often speak about agency as though it were purely a personal attitude.
Be curious.
Take initiative.
Believe in yourself.
Start building.
There is value in this language.
But it can also disguise structural conditions.
It is easier to experiment when failure will not threaten your livelihood.
It is easier to challenge assumptions when your position is secure.
It is easier to start a project when you have spare time.
It is easier to take entrepreneurial risk when you have financial support.
It is easier to recognise opportunities when you have strong networks.
It is easier to question an AI answer when you have enough education to recognise that something is wrong.
It is easier to exercise voice when institutions are prepared to listen.
Agency is therefore both personal and contextual.
| Individual resources | Environmental resources |
|---|---|
| Curiosity | Permission to experiment |
| Confidence | Psychological safety |
| Knowledge | Access to expertise |
| Judgement | Quality information |
| Time | Manageable workload |
| Networks | Connection to opportunity |
| Financial resilience | Room to fail |
| Initiative | Institutions responsive to initiative |
Telling everyone to “become a creator” without acknowledging these differences risks turning structural inequality into an individual character judgement.
Some people are not waiting for permission because they lack imagination.
They may have learned, quite rationally, that acting without permission carries consequences.
The compounding advantage of AI
This is where AI becomes particularly interesting.
Technologies that increase capability can have multiplicative effects.
Consider someone who already possesses:
- substantial domain expertise;
- excellent communication skills;
- strong professional networks;
- financial resources;
- confidence;
- institutional authority; and
- time to experiment.
AI does not merely add one more capability.
It potentially amplifies all the capabilities they already have.
Their expertise helps them prompt better.
Their knowledge helps them evaluate the output.
Their networks help them distribute what they create.
Their authority helps them implement it.
Their resources allow them to experiment repeatedly.
Their confidence encourages them to try.
Now consider someone without those advantages.
They may have exactly the same subscription.
But access to the model does not magically create the surrounding ecosystem required to convert an idea into action.
This is why the simplistic claim that AI “levels the playing field” deserves scrutiny.
It may level some parts of the field while making other differences more important.
Perhaps expertise becomes leverage
There is another paradox.
AI is often presented as a substitute for expertise.
Sometimes it clearly is.
A novice can now accomplish tasks that previously required specialist knowledge.
But expertise may simultaneously become more leveraged.
Suppose AI allows both an expert and a novice to perform a task five times faster.
The novice becomes more capable.
That is valuable.
But the expert now combines fivefold productivity with years of accumulated judgement, context and networks.
The technology may therefore narrow one gap while widening another.
The scarce resource moves again.
Not merely:
Who can produce?
But:
Who knows what to produce?
Who can evaluate it?
Who can convince others?
Who has authority to implement it?
Who owns the resulting value?
These are questions of agency and power.
Education sits directly inside this problem
This is where universities have an important responsibility.
If AI capability becomes widely available, higher education should not simply teach students how to operate AI tools.
That would address only the shallowest layer of the divide.
Students need opportunities to develop the deeper resources through which capability becomes agency.
They need to practise:
- asking their own questions;
- making decisions under uncertainty;
- challenging authority constructively;
- collaborating with people unlike themselves;
- testing ideas with real users;
- recovering from failure;
- defending a position;
- changing their mind;
- developing professional networks;
- communicating ideas to people with power;
- recognising when AI is wrong; and
- imagining themselves as people capable of changing something.
This is not simply an AI literacy curriculum.
It is an agency curriculum.
From teaching tools to creating conditions
That distinction matters.
Universities could respond to AI by creating compulsory modules:
Introduction to Generative AI.
Prompt Engineering 101.
Responsible AI Use.
Some version of these may be useful.
But imagine two graduates.
Both understand prompting.
Both know how LLMs work.
Both can explain hallucination and bias.
One has spent three years predominantly responding to carefully structured assignments.
The other has repeatedly worked on ambiguous problems, collaborated with external partners, initiated projects, made decisions, failed, reflected and tried again.
Which graduate is more prepared for an environment in which AI makes technical capability abundant?
Probably the second.
Not because they know more prompts.
Because they have practised acting.
Agency should not mean individualism
There is another distinction worth making.
The language of agency can slide easily into heroic individualism.
The lone entrepreneur.
The visionary founder.
The person who ignores everyone else and “just ships”.
That is not the model I am advocating.
Many of the problems most worth solving are collective.
Climate transition.
Public health.
Ageing.
Education.
Housing.
Inequality.
Trust.
Community cohesion.
AI itself.
No individual creator solves these alone.
Agency in these contexts means the capacity to contribute meaningfully to collective action.
To convene.
To collaborate.
To build trust.
To understand systems.
To recognise power.
To work across boundaries.
To help groups move from discussion towards action.
Perhaps one of the most important forms of agency in the AI era will not be:
“I built this.”
but:
“We made this possible.”
An institutional agency divide
There is also a larger version of the same problem.
Universities themselves differ enormously in agency.
Some institutions have:
- capital;
- computing infrastructure;
- strong industry relationships;
- large research budgets;
- specialist AI teams;
- access to proprietary systems; and
- considerable negotiating power.
Others will largely purchase whichever AI products vendors make available.
This mirrors an argument from After Scarcity.
Universities helped create the knowledge foundations of contemporary AI, yet risk returning to the resulting ecosystem primarily as customers.
Agency therefore matters at multiple levels.
Individual agency.
Organisational agency.
National agency.
Who merely consumes intelligence?
Who shapes how it is used?
Who owns the infrastructure?
Who sets the rules?
Who captures the value?
Who gets to choose?
The agency divide may eventually prove as important as the digital divide.
The university as an agency institution
This brings the series back to its starting point.
If intelligence becomes abundant, what exactly is a university for?
Perhaps part of the answer is this:
Universities can help people become authors of their own participation in the world.
Not authors in the narrow sense of individual autonomy.
People who understand enough to make informed choices.
Who can recognise possibilities.
Who can evaluate competing claims.
Who can create.
Who can collaborate.
Who can challenge.
Who can take responsibility.
Who can act.
That is a more ambitious goal than employability.
But perhaps it is also a more durable one.
The jobs students enter will change.
The technologies will certainly change.
Today's AI systems will look primitive surprisingly quickly.
Specific technical skills will come and go.
Agency sits beneath them.
After Intelligence
This series began with a question:
What kind of person becomes valuable when sophisticated capability becomes abundant?
Across four essays, the answer has moved through four ideas.
Agency
When capability becomes accessible, the advantage shifts towards recognising problems and initiating action.
Creation
When the cost of making things collapses, making something is no longer sufficient.
Discernment
When almost anything can be generated, selecting what is good becomes more valuable.
Agency inequality
And when everyone has access to similar capabilities, differences in confidence, knowledge, networks, opportunity and power may determine who actually benefits.
Which gives us a progression:
Capability → Agency → Creation → Discernment → Contribution
But perhaps there is one idea underneath all four.
AI gives us leverage.
It does not tell us where that leverage should be applied.
It does not guarantee that everyone gets the same opportunity to apply it.
And it does not absolve us of responsibility for what happens when we do.
The question after intelligence
For the past few years we have repeatedly asked:
What can AI do?
That was the right question at the beginning.
Increasingly, it is not enough.
We need to ask:
What can we now do because AI exists?
And then:
What should we do?
And finally:
Who gets the opportunity to decide?
Those are questions technology cannot answer for us.
They are questions about education.
Institutions.
Power.
Values.
Judgement.
And ultimately about what kind of humans we want to become alongside increasingly capable machines.
Perhaps that is the deeper challenge after intelligence becomes abundant.
Not learning how to compete with the machine.
Not simply learning how to operate it.
But developing enough agency, judgement and purpose to decide where we want all this new capability to take us.
The complete series
After Intelligence: From Capability to Agency in the Age of AI
- Agency — Universities Prepared Us to Find Work. What If the Future Is About Creating It?
- Creation — When Everyone Can Build Anything, Knowing What to Build Becomes Everything
- Discernment — AI Can Make Almost Anything. Can We Still Tell What Is Good?
- Agency Inequality — AI Gives Everyone More Capability. It May Not Give Everyone More Agency.
References
Farmer, I. (2026). After Scarcity: Rethinking the University in the Age of AI. Frontiering.