What happens to a society when machines can do more and more?

Society rarely changes politely and in an orderly way

When a new technology gets big enough, it doesn’t stay in the IT department. It slips into schools, companies, public authorities, relationships, job profiles and eventually our image of what it means to be human.

That is where we stand with AI.

You can describe it technically at first. Faster analysis. Automation. New tools. Language models. Agents. Systems that can see, hear, write, structure and act.

But it quickly becomes a societal question. Because as soon as machines take over more of what we long considered typically human mental work, we have to clarify anew what we humans actually still want to do ourselves, should do ourselves and have to answer for ourselves.

And with that we are right in the middle of human intelligence.

Not as a friendly opponent to artificial intelligence. But as a very practical question: which human abilities become more important rather than less important in an AI society?

The world of work and education

It is most visible in the world of work first. When AI writes texts, prepares analyses, helps develop software, sorts support cases or speeds up routine decisions, it is not just activities that disappear. Above all, what people are still needed for in daily work shifts.

Many jobs are therefore not abolished but rebuilt. And in some ways that is more demanding. Because then it isn’t enough to introduce a piece of software once and carry on as before. People have to learn to check results, steer systems, spot errors, take responsibility and step in where formally plausible answers no longer match reality.

That is exactly why education is not a side issue here.

When work changes, education has to change with it. Not only in the sense of: everyone now learns a few new tools. But much more fundamentally. What should people be able to do when knowledge is available at any time, standard texts appear in seconds and even analytical groundwork is increasingly done by machines?

Perhaps what is hardest to automate becomes more important: understanding connections. Asking good questions. Forming judgements. Dealing with uncertainty. Thinking with others. Being able to correct yourself.

In the world of work and education, human intelligence is therefore not a remnant of old humanity but the ability to cope with a world in which pure information processing is no longer automatically our unique selling point.

Justice and relationships

AI never hits a society evenly.

Some people have access to good tools, good education, good infrastructure and enough time to get used to new ways of working. Others don’t. Some benefit from automation. Others are mainly monitored more closely, standardised or pushed out of processes that at least used to have a human counterpart.

That is why, with AI, justice does not become a moral footnote but a rather hard design question.

Who gets to benefit from the productivity gains? Who bears the risk of wrong decisions? Who can defend themselves against automated assessments? Who even still understands by what logic decisions about them are made?

And then there are relationships.

That sounds softer at first, but it is no smaller for society. As machines get better and better at listening to us, answering, mirroring, advising and sometimes even comforting, that inevitably changes the space between people. A machine doesn’t suddenly become a person. But people start to shift parts of their attention, trust and closeness into technical systems.

I consider that neither pure dystopia nor harmless.

It is simply something we should take seriously. Because a society doesn’t live only on working processes. It lives on people noticing one another, contradicting one another, putting up with one another and not replacing one another entirely with efficient interfaces.

Here, human intelligence means above all social intelligence: being able to tell closeness from simulation, not trading vulnerability for convenience and noticing when digital answers merely imitate the interpersonal.

Ethics and governance

In the end it all lands at ethics and governance, even if both terms unfortunately quickly sound like a panel discussion.

In this context, ethics is nothing lofty. It is not about solemnly declaring values at conferences. It is about who gets to decide by what standards systems act, whose interests are written into models and what happens when technical efficiency collides with human fairness.

Governance is the practical side of that.

Who controls these systems? Who is liable? Who may use which data? Which limits apply in companies, in schools, in administrations, in hospitals? And how do we prevent responsibility from always elegantly disappearing where, supposedly, only the system issued a recommendation in the end?

It is precisely here that an AI society needs significantly more human intelligence, not less.

It needs people who ask not only what works technically but also what is wise for society. People who say what is uncomfortable. People who organise dissent. People who, amid all the enthusiasm for automation, don’t forget that democratic, social and moral orders are deliberately more complicated than a well-optimised process.

Perhaps that is the decisive point anyway.

AI doesn’t just force us to think about machines. It forces us to do something much more uncomfortable: to get clear about what we actually mean by good work, good education, fair participation, real relationships and responsible steering.

And that is exactly why this is not a fringe topic for philosophers, ethics councils or digital policy makers.

It is a very practical question for all of us.