Who are we becoming because of AI?
In brief: New major technologies tend to radically transform the societies they emerge within, but AI is the first one beginning to act on its own. If history is any guide, its effects will reach far beyond which jobs disappear or what the technology itself can do. AI could reshape how wealth is created and distributed, what expertise is worth, who holds power, and perhaps even how we understand work and our own value within society. The transition is unlikely to be smooth, and the choices we make about who owns and governs AI, and what we ultimately use its capabilities for, will help determine what emerges from it. Most of our attention is going to the question of what AI will become. This essay asks the other one: who are we becoming because of AI, and could this moment of disruption be used to build a society in which people and the wider living world can flourish?
I use AI most days now, both for work that would once have taken me far longer and for personal use. I use it to find my way into unfamiliar research, test an argument, pull together material I might otherwise struggle to locate, and, often, to learn and simply help me think.
I also find myself resisting it as there are things I do not want it to do for me because I suspect the difficulty is part of the point. Sometimes I need to sit with an idea for a while, fail to articulate it properly, come back to it, realise that the thing I thought I was trying to say was not actually the thing at all. An AI can shorten that process considerably but I am not always sure shortening it is good for me.
That ambivalence seems fairly common. Some people I know are vehemently opposed to AI, for many good reasons. They worry about artists and writers whose work has been absorbed into training data, about energy use, surveillance, loss of jobs, the enormous concentration of power around a handful of companies etc. Others find it exhilarating and are experiencing something genuinely remarkable: suddenly they can code, research, translate, produce images, analyse information or access forms of expertise that were expensive or inaccessible only a short time ago.
Most of us, I suspect, sit somewhere between those positions, using AI while wondering what exactly we are becoming involved in.
I think that all three positions, whilst quite different one from another other reflect some form of common sense. We are all trying to make sense of a transformation while already participating in it, whether through our own use of AI or through workplaces, governments, businesses and information systems increasingly using it around us. So it is normal that we have different yet sensible opinions about it. Our opinions will probably change too, because AI itself keeps changing.
Much of the conversation has understandably focused on where the technology is heading. How capable will these systems become? What jobs will they replace? When might we reach artificial general intelligence, if we reach it at all?
I find myself increasingly interested in a different question: who are we becoming because of AI?
(Picture by Aerps)
A technology that can act
Previous information technologies helped humans communicate, travel, transport, store knowledge and exercise power. The printing press could spread an argument across a continent… but somebody first had to think the argument. A telephone could connect two people thousands of kilometres apart… but the communication was happening between two humans.
The historian Yuval Noah Harari has drawn attention to something about AI that helps explain why the question matters. He argues that AI is different because it is beginning to acquire agency.
We can now give an AI system an objective and have it determine at least some of the steps required to achieve it, searching for information, reasoning through problems, writing code, communicating with humans and other systems, adjusting what it does according to the results it encounters. These capabilities remain imperfect and frequently unreliable, but the direction is towards greater agency.
In other words, we have, for the first time in history, introduced into our societies technologies that can themselves become actors within those societies.
I am not thinking here in whether that ultimately produces the science-fiction versions of AI we have been imagining for decades (Terminator anyone?) than in what happens well before then, when millions of artificial agents begin operating inside businesses, schools, government departments, financial markets, political campaigns and ordinary households.
Indeed, technologies do not simply arrive and sit politely inside the societies that created them. They change the societies.
When a technology changes the system around it
The history of major technological change suggests that its deepest consequences often appear well beyond the technology.
For instance, it is true that the railway made transport faster. But this is an incomplete picture of what happened. Railways changed where towns could exist, what farmers could sell and how far away from a market they could live. They altered the relationship between cities and their hinterlands. They helped standardise time itself because a national rail system became difficult to operate when every town kept its own local time.
The car did something similar. We did not simply replace horses with engines but redesigned cities around roads and parking. Suburbs moved further from workplaces. Shops changed location. Oil became entwined with national security and families organised parts of their lives around the assumption that an adult could drive somewhere whenever they needed to.
Well before that, the printing press changed the economics of producing and distributing knowledge, weakening established monopolies over information and becoming entangled with the Reformation and the enormous religious and political upheavals that followed.
The Industrial Revolution went deeper still. Not only steam power, mechanisation and railways transformed production and movement, but their consequences extended much further. Industrial cities grew, populations moved, fortunes were accumulated on an extraordinary scale and new forms of exploitation developed alongside them. Relationships between workers and owners changed, as did the political power available to each.
So, people responded to the society emerging around them. Trade unions developed, and new political theories and systems fought for relevance. Socialist and communist movements grew, anarchists proposed radically different ways of organising society, while liberalism and conservatism evolved in response to conditions their earlier proponents could barely have imagined. Over time, and through considerable conflict, came labour protections, public education, social insurance and welfare states.
This is roughly what scholars of socio-technical transitions are getting at: a major technology becomes entangled with institutions, economic interests, law and everyday practice. Those elements change one another until the system itself starts to look different, and AI seems likely to do this across several systems at once.
When intelligence gets cheap
For most of modern history, sophisticated cognitive capability has been scarce and expensive. If we needed legal analysis, engineering advice, software, translation, financial modelling or medical expertise, we generally needed another human being who had spent years acquiring the knowledge required to provide it.
AI is beginning to alter that scarcity.
Already, one person can access capabilities that would recently have required several people or specialised services. A small organisation can analyse information it could never have afforded to analyse. A child can have access to a tutor able to respond endlessly - and patiently - to their questions. Scientists can interrogate vast datasets and explore possibilities at scales beyond unaided human cognition.
The potential gains are immense. AI could accelerate medical research, help us understand climate systems, design new materials, optimise energy networks and help organisations and governments model complex risks. It may remove vast amounts of administrative work that consumes people's lives while contributing relatively little to them. For people without money to hire consultants, tutors, programmers, lawyers or other specialists, access to reasonably capable machine intelligence could redistribute forms of capability that have long followed wealth.
But then those gains start interacting with the rest of society. If AI can perform much of the work currently done by a junior lawyer, accountant, designer, programmer or analyst, organisations may eventually need fewer people in those positions. The immediate economic consequence falls on the person whose job disappears, but another problem follows later because today's junior practitioner is how we produce tomorrow's experienced practitioner. Remove enough of the bottom rungs and eventually something happens to the ladder.
Education is affected too. Much of our educational system developed in a world where access to knowledge was scarce, acquiring expertise required sustained effort and the ability to produce a sophisticated essay or solve a difficult problem provided some evidence that a student understood the material. When an AI can perform those tasks within seconds, some of the methods we have used for learning aren’t as relevant. We may have to reconsider what education is trying to cultivate, including what kinds of human capability children need when competent machine intelligence is available everywhere.
The economic consequences could travel much further. If organisations can produce considerably more value with fewer employees, wealth can grow while the relationship between employment and prosperity weakens. Yet we have built an extraordinary amount around that relationship. Most people obtain income through work, governments obtain substantial revenue by taxing that income, mortgages depend upon its continuity and welfare systems generally assume employment as the normal condition from which people occasionally fall away.
A large enough change to the role of human labour therefore travels far beyond the labour market.
Then there is meaning
For many of us, work is woven deeply into identity. We ask someone we have just met what they do and understand the answer as telling us something about who they are. We spend years becoming competent in a field, and the difficulty of acquiring that competence is part of what gives it value, both to society and to ourselves. Expertise can bring status and belonging, a place within a community of practice, the experience of being needed by others and the satisfaction of becoming capable at something that was once difficult.
This relationship has always been complicated. Work can exploit and exhaust us, many jobs provide little meaning at all, and some of the work AI may remove is precisely the work people would gladly surrender. There is real liberation in that possibility. Keynes imagined nearly a century ago that technological progress might eventually allow his grandchildren to work around fifteen hours a week. Instead, we became extraordinarily more productive and found endless new things to do. AI could confront us with that choice again.
There is also something worth taking seriously in what may happen when capabilities around which people have organised substantial parts of their lives suddenly become cheap. A translator who has spent decades learning the subtleties of language, a graphic designer who developed their craft over thousands of hours, a researcher who learned how to navigate a body of literature, a programmer who knows how to build something from nothing, may experience AI as an expansion of their abilities and simultaneously as a destabilisation of the value attached to those abilities.
For younger people, the shift may run deeper because they may never experience some cognitive tasks as things humans normally do unaided. Many of us no longer remember telephone numbers because our devices remember them for us. What happens when the outsourced capacities are writing, research, navigation through complex information, memory and increasingly reasoning itself?
Some of that may free cognitive space for other forms of intelligence, some may diminish capabilities we later discover we needed.
There is no obvious line separating augmentation from dependency, and the line will not sit in the same place for everyone. What interests me is that this transition may force us to reconsider a relationship we have largely taken for granted between intelligence, competence, work, social value and our sense of ourselves.
If intelligence itself becomes abundant, we may eventually have to think differently about what we value in one another.
The transition will be difficult
Technological futures are often presented by comparing the present with an imagined destination. Perhaps AI eventually creates extraordinary material abundance, scientific breakthroughs and much shorter working weeks. Perhaps education becomes individually tailored and high-quality expertise becomes available to almost everyone.
Some version of that future is conceivable, but history gives us reason to pay closer attention to how we get there.
Industrialisation eventually contributed to enormous improvements in material living standards, although those gains were neither immediate nor evenly shared. People were displaced from older forms of livelihood and drawn into rapidly expanding industrial cities where housing, sanitation and public institutions struggled to keep pace. Factory workers, including children, worked under conditions we would now find abhorrent, while owners accumulated fortunes on a scale that transformed political power as well as economic life.
The resulting conflicts helped shape what industrial society eventually became. Workers organised because the distribution of the gains from industrialisation was contested. Governments regulated workplaces because unregulated industrial relations produced conditions societies became increasingly unwilling to tolerate. Welfare institutions developed partly because the market economy produced forms of insecurity that families and communities could no longer absorb alone.
Many institutions we now associate with relatively stable industrial democracies were built through decades of conflict over who would receive the benefits of a new productive system and who would carry its costs.
Something similar may confront us with AI, although potentially much faster.
Consider the professions. If organisations automate large amounts of entry-level cognitive work, the immediate benefit accrues largely to organisations and their owners through reduced costs and increased productivity. The immediate burden falls disproportionately on people trying to enter those professions, including young adults who may have invested years and considerable debt acquiring qualifications for a labour market changing beneath them. Eventually the organisations themselves may discover that they have weakened the pathways through which experienced human judgement was previously developed.
Or consider public finance. A society in which corporations produce increasing value using decreasing amounts of human labour could become wealthier in aggregate while simultaneously eroding one of the principal mechanisms through which that wealth currently reaches households and governments. If labour income falls relative to returns on capital, the question of who owns the productive systems becomes increasingly consequential. Universal basic income, public AI infrastructure, shorter working weeks, taxation of machine-generated wealth and different forms of collective ownership then cease to be curiosities of technological futurism and become real arguments about who gets what from the new economy.
Our information environment may be even more exposed. Generative AI drastically reduces the cost of producing plausible text, images, audio and video, while agentic systems may reduce the cost of distributing and continuously adapting that material to particular audiences. The consequences extend beyond misinformation in the familiar sense. Political persuasion can become personalised and persistent, delivered through interactions that increasingly resemble ordinary human conversation. In societies where trust in governments, media and other institutions is already fragile, establishing a sufficiently shared account of what has happened may itself become harder.
The environmental consequences expose another part of the problem. Much attention has understandably focused on the electricity and water consumed by AI data centres, but recent research suggests that the indirect effects may matter just as much. A modelling study reported by The Guardian examined what happens when AI increases productivity across the energy sector. Across 64 scenarios, researchers estimated that AI-enabled gains in coal, oil and gas production could produce an additional 0.47 to 1.8 gigatonnes of carbon pollution each year, around 1 to 5 per cent of annual energy-sector emissions. In other words, while AI can help renewable operators reduce downtime and manage electricity grids, it also helps fossil-fuel companies locate reserves, reduce drilling costs and bring projects online faster. Saudi Aramco says it has embedded AI throughout its operations, while Equinor has credited AI and new seismic technologies with helping make discoveries on the Norwegian continental shelf.
The social effects will similarly be experienced unevenly.
Losing a job can mean losing income, but it can also mean losing status, relationships, a daily structure and the experience of being useful. When this happens across a community, the effects spread. We have seen versions of this in places affected by deindustrialisation, where the closure of a factory or mine removed much more than employment. Local businesses disappeared, younger people left, civic institutions weakened, families absorbed the strain and resentment accumulated around the sense that decisions were being made elsewhere by people insulated from their consequences.
Private experiences of insecurity will then likely acquire a political life as people will want an explanation for why their life has become less secure and why somebody else seems to be doing extremely well out of the change. Political movements provide answers. Sometimes that produces solidarity, organisation and demands for a fairer settlement, but it can also produce scapegoating, exclusion, violence and movements promising to restore a social order that has already disappeared.
AI will create different geographies of disruption, reaching into professions and communities that have regarded themselves as relatively protected from automation. And this will happen alongside climate disruption, housing pressures, geopolitical instability, inequality and declining institutional trust. Shocks accumulate and eventually they become political.
This is one reason I expect the coming years to produce a flurry of political ideas around AI, including ideas that currently seem implausible. Industrial capitalism helped generate Marxism, organised labour, social democracy and eventually welfare capitalism because societies needed ways of making sense of, contesting and governing a radically altered economic order.
We should expect our own versions of that intellectual and political ferment. Universal basic income is already part of the discussion, but it is unlikely to be the last word. We may see serious proposals for taxing machine-generated wealth or treating access to AI capability as something like a public entitlement. Shorter working weeks may return as a central labour demand. There will probably be ideas that currently sound bizarre and later become normal.
Who owns the intelligence?
Greater intelligence does not arrive with an ecological or social preference built into it. It increases capability, including the capability to continue doing things that are already causing us harm.
So before asking what kind of future we want from AI, we probably need to spend more time on a more immediate question: who gets to control it?
The US model currently developing at the technological frontier gives us one answer. Extraordinary capability is being built largely through private corporations and staggering concentrations of capital, computing infrastructure and energy. OpenAI's partnership with Amazon alone includes a US$50 billion Amazon investment and an expanded US$100 billion infrastructure agreement over eight years, involving around two gigawatts of computing capacity.
There is a certain logic to this. Frontier AI is extraordinarily expensive to develop and the American technology ecosystem has proved remarkably effective at mobilising private capital, attracting talent, taking risks and producing rapid innovation. We are receiving capabilities at a pace that publicly managed technology programs would struggle to match.
We are also allowing a remarkably small group of corporations to build infrastructure that may eventually mediate knowledge, work, education, commerce and significant parts of public life.
China offers a different political economy of AI. Chinese technology companies remain major actors, including highly innovative ones such as DeepSeek, while the state plays a much more directive role in setting industrial priorities, building infrastructure and determining the political boundaries within which these systems operate. China has increasingly framed AI internationally as something that should benefit humanity broadly, explicitly calling it an international public good, advocating greater access for developing countries and establishing a new World Artificial Intelligence Cooperation Organization in 2026.
There is something worth examining in that claim, but without romanticising the Chinese model. State direction does not make technology democratic as government control of powerful AI systems creates its own risks around surveillance, censorship, propaganda and the concentration of political power. Nor is China's AI economy somehow outside markets or corporate competition.
Still, China raises a question that Western countries have largely left to the market: if machine intelligence becomes foundational infrastructure, why should we assume that its most capable forms will remain primarily private property?
Perhaps some AI capability will eventually come to resemble roads, electricity grids, libraries, universities or the internet itself, infrastructure whose benefits are so widely distributed through society that purely private provision becomes politically difficult to justify.
That opens questions about public models, publicly funded compute, open-source systems, cooperatives, sovereign AI infrastructure and international institutions capable of providing poorer countries access to capabilities they could never afford to develop themselves.
It also opens a harder question about governance. If AI agents eventually participate extensively in our economies, information systems and public institutions, who decides the rules under which they operate? Governments? The companies that created them? International bodies? Users themselves?
We have barely begun that argument, and it will ultimately be an argument about power as much as technology.
Transition towards what?
Even a well-governed AI system can help us pursue foolish goals more efficiently.
Our societies already produce remarkable wealth while degrading many of the ecological systems upon which that wealth depends. We extract materials, burn fossil fuels and then measure much of the resulting throughput as progress.
AI is entering that system as an accelerator and the fossil-fuel example makes the point rather brutally. AI can help us find more oil and it can also help us manage a renewable electricity grid. If those goals continue to revolve around increasing consumption and extraction on a finite planet, greater intelligence may accelerate some of the dynamics already destabilising the conditions upon which our societies depend.
The algorithm has no view about which civilisation it would prefer to support. We decide that through economics and politics.
Three Horizons
One way I have found useful for thinking about this comes from a futures framework called Three Horizons.
The first horizon describes the dominant system we have inherited, including the institutions and practices that continue to organise much of our lives even as some become increasingly poorly fitted to emerging conditions. AI is currently being absorbed into this world, which is why so much innovation is directed towards familiar objectives such as increasing productivity, reducing costs and expanding markets.
The second horizon is messier. The existing system remains dominant while experiments proliferate around its edges and increasingly within it. Some make the first horizon more efficient and extend its life while others begin creating practices from which a different system might eventually grow. Some will fail, some will be captured by existing interests and some may start out looking insignificant and later prove transformative. This is probably where much of the struggle around AI will take place.
Then there is the third horizon, which asks us to consider what we would actually want to emerge.
This is where regenerative thinking gives me something I find missing from much of the AI conversation. It asks us to look beyond whether an activity generates more output and consider whether human activity contributes to the capacity of the wider living systems of which we are part to flourish. Applied to AI, that changes the direction we’re taking.
We could use AI to reduce the materials and energy required to meet human needs, help restore forests, soils, oceans and rivers, understand ecosystems that are extraordinarily difficult for us to model, accelerate the transition away from fossil fuels and make sophisticated knowledge available to communities that currently cannot afford it.
We could also do something much simpler with the productivity dividend and simply work less. If a task that once took ten hours takes five, the five hours that have been released have to go somewhere. They can become more output, another client, more consumption and another increment of GDP. They can also become time with our children, care for someone, participation in community, making something with our hands, walking through a forest, thinking without producing anything, resting. That is also a productivity dividend, although our economic statistics struggle to see it.
I do not think "regenerative" gives us a ready-made future, but it pushes us beyond the assumption that the purpose of intelligence is to maximise whatever our current institutions already reward. We could ask instead whether a technological system leaves societies with more agency or less; whether the productivity it creates becomes wealth concentrated at the top or time returned to people lower down; or whether it restores ecological capacity or increases extraction.
Those choices will rarely be clean as societies might welcome AI because it gives a small local organisation access to expertise it could never afford, while opposing the data centre proposed upstream because of its water use. A worker may resent automation that threatens their job and still use the same technology at home to help their child learn maths.
What does this mean for those of us using AI today?
I use AI extensively, including while developing this article. I use it to explore research, interrogate ideas, find connections I had missed and push against arguments that feel incomplete. It expands what I can do as one person, sometimes dramatically, and I find that genuinely exciting.
But I also notice the temptation to hand over things I would once have struggled with myself. Sometimes it is entirely sensible as I have no particular desire to preserve hours of administrative work simply because humans used to perform it. At other times the struggle is part of the work. As I mentioned earlier, pondering an idea that has not yet resolved itself, finding language for something I can sense but cannot articulate, reading something difficult several times until I understand it, these experiences develop capacities that matter to me independently of how efficiently I produce an output.
So I am still working out where that boundary sits for me, but there is only so much any of us can determine individually.
Decisions being made by governments and technology companies will shape this transition at a scale vastly beyond our personal choices. Yet organisations decide how they distribute productivity gains. Societies decide which technologies they incorporate into collective life. Schools decide what children still need to learn to do themselves. We decide which capacities we are comfortable outsourcing and which we continue exercising because they are part of the people we want to become.
Which brings me back to the larger question: who are we becoming because of AI? We do not know what AI will ultimately become, but this period of technological disruption gives us reason to reconsider some questions we should probably have been asking anyway.
What is an economy for?
What should work give us beyond a wage?
If machines produce more of what we need with less human labour, who receives the benefit?
What capabilities do we want our children to develop for themselves, even where a machine can perform them faster?
How much is enough?
Who should own an intelligence that becomes part of the infrastructure of society?
And on a planet whose living systems are already under immense pressure, what would it mean to organise our human and increasingly non-human intelligence around enabling life to flourish?
I don’t know what answers we will collectively reach, but I am increasingly convinced that these are better questions than asking only what the next model will be able to do.
We spend enormous intellectual energy asking what AI will become. Perhaps we should spend more of it asking what we will become because of AI.
* * *
If this article resonated with you, I invite you to check out an earlier article of mine which provides practical guidance for organisations that want to engage with AI deliberately rather than reactively.
Need help applying these ideas inside your organisation? I work with boards and executive teams to build crisis-ready capability across leadership, culture and systems.