When Adoption Outpaces Society’s Capacity

These figures show just how quickly a technology can move from something we talk about as the future to something already embedded in everyday life, often without us quite noticing when the shift took place. (The Swedes and the Internet 2026)

In many ways, the development is impressive. People use AI to write, summarise, translate, code, search for information, understand complex issues and tackle tasks that would previously have required significantly more time or specialised expertise. More than eight in ten students and seven in ten white-collar workers use AI, and 56 per cent of all employed people say they have used AI at work. At the same time, only around half feel that the technology actually makes them more efficient.

Perhaps that is where the report’s most interesting observation lies. Not in the rapid increase itself, but in the gap between adoption and actual capability. Because it is entirely possible to introduce a technology without having built the capacity to use it well. Organisations can buy licences, provide access to tools and encourage experimentation without creating the skills, working practices and governance needed for the technology to contribute to better decisions, higher quality or greater value for the people the organisation exists to serve. When adoption develops faster than the organisation’s capacity, responsibility easily falls on the individual employee, who is expected to decide what is appropriate to share, which answers can be trusted, which risks need to be managed and when an apparently efficient way of working is actually creating new dependencies or shifting costs elsewhere in the system.

When Every Decision Is Rational

Much of digital development looks exactly like this. Each individual decision may be perfectly understandable from the position in which it is made. An employee saves time by using an AI tool, a manager wants to increase productivity, an organisation chooses the service that is quickest to implement and easiest to use, and a supplier offers a solution that responds to demand already present in the market.

Yet the sum of all these rational decisions can lead to an outcome that nobody actually chose. We can build long-term dependencies on a small number of global technology companies, make public services dependent on infrastructure that we neither own nor fully understand, and allow sensitive assessments to move into systems whose logic is difficult to scrutinise. At the same time, expertise that once existed within the organisation may gradually erode, because in the short term it appears both cheaper and easier to buy a ready-made service.

I have written about this before in the essay When Every Decision Is Rational and the Outcome Is Still Wrong, where I explore how systemic failures can emerge without any individual actor having made an obviously bad decision. AI makes that question even more urgent, because change is now happening at a pace that allows adoption to become established long before organisations have had time to develop shared assessments, rules and structures. This does not mean that the answer is to try to stop technological development, nor would that probably be possible. But it does mean that we need to stop confusing rapid adoption with successful digital transformation.

Digitalisation Is More Than the Spread of Technology

For a long time, we have talked about digitalisation as though the main challenge were getting people and organisations to start using technology. From that perspective, high adoption becomes a sign of success and caution something to be overcome. But when a technology is already being used by a majority of the population, the question changes. We need to be able to assess what that use actually leads to, what value it creates, which groups benefit and which risk being left behind, what new vulnerabilities emerge and who carries responsibility when the consequences no longer fit within the boundaries of a single organisation.

The Swedes and the Internet report shows that eight in ten Swedes regularly find themselves stuck in passive scrolling, and that 44 per cent say it makes them feel worse, while only two per cent say it makes them feel better. Half are interested in taking a break from digital life. It is an important reminder that a digital service can be successful when measured by usage while simultaneously creating consequences that users themselves are left to manage. Technology can be easy to start using, difficult to put down and profitable for the company providing it without necessarily contributing to the society we want to build.

A mature digitalisation policy therefore needs to hold several perspectives at once. We need to seize the opportunities technology offers without turning people’s capacity to adapt into our primary form of risk management. We need to create room for innovation while also building shared capacity in information security, law, procurement, data governance, accessibility and ethical assessment. Above all, we need to measure more than the number of users and tools introduced, because the crucial question is whether technology contributes to the organisation’s purpose and to a society that works better for people.

Who Gets the Mandate to Build the Future?

As technology moves into the core of organisations, the expertise required to lead that development also changes. It is not enough to understand what systems are technically capable of doing. We also need to understand the organisations in which they will be used, the people affected by them, the dependencies they create and the conflicts between competing goals that cannot be solved by adding another technical feature. Yet digitalisation is still often organised as though it primarily belonged within an IT function, separate from the organisation’s core activities and at a comfortable distance from the rest of the executive team’s decisions. As a result, we risk treating questions of responsibility, power, working conditions, security, expertise and societal impact as technical implementation issues when they are, in fact, part of the organisation’s strategic direction.

In Who Gets to Build the Future?, I have written about how our understanding of technical expertise shapes which people, experiences and perspectives are given status and authority in transformation. When the ability to connect perspectives, understand dependencies and formulate problems across organisational and disciplinary boundaries becomes increasingly important, while organisations continue to reward a much narrower conception of technical expertise, an uncomfortable lag emerges. We need one kind of capability, but continue to give authority to another.

This is not a side issue in the development of AI. Who we choose as leaders, the experiences they bring with them and the values they embed in their organisations will have a major influence on how the technology is used. If leadership primarily rewards speed, implementation and measurable productivity, other values can easily disappear from decisions, not because anyone has actively chosen to exclude them, but because they were never given equal weight in the first place.

When the Market’s Direction Becomes Society’s Direction

In a widely discussed conversation at Colgate University recently, Barack Obama described the development of AI in a way that captures the contradictory situation we now find ourselves in. In his view, the technology itself is not overhyped, even if the commercial promises surrounding it and the valuations of AI companies may well be. It is an important distinction, because it allows us to take the technology’s actual power seriously without having to accept every forecast, business model or story about the future that has been built around it.

Obama describes how development accelerates as models become increasingly capable of contributing to their own continued development. But the most interesting part of his argument is not really the prospect of some future superintelligence. It is his distinction between two different kinds of alignment problem. The first is the one commonly discussed in AI safety: the risk that technical systems begin acting in ways that do not align with human intentions. The second is more immediate, already visible and in many ways more political: the gap between how technology could best be used to serve society’s needs and the commercial pressures shaping the companies developing it.

The companies at the forefront of AI development have raised enormous amounts of capital. Investors expect returns, companies need to create products that people and organisations are willing to pay for, and the technology therefore has to reach the market at a pace that justifies those investments. This does not mean that these companies are run by people who wish society harm, any more than the individual organisations buying their services are making obviously irrational decisions. But it does mean that the direction of a technology rapidly becoming embedded in education, working life, public administration and people’s most private spaces is being shaped to a considerable extent by commercial incentives, even though its consequences are shared by society as a whole.

Obama concludes that voluntary restraint by companies may be necessary in an acute situation, but that it can never replace effective public governance. The market cannot be expected to manage on its own the risks associated with systems that affect people’s safety, livelihoods, access to information and ability to participate in society, just as we have not left the entire responsibility for pharmaceuticals, food or aviation safety to the companies selling those products. (Watch the conversation with Barack Obama)

This brings us back to the privilege of defining the problem. When the same companies that develop the technology, own the infrastructure and profit from its use are also given considerable scope to describe the risks and define which responses are possible, they influence not only the market but also our understanding of what the problem actually is. If AI development is primarily framed as an international race, faster development becomes the obvious solution. If the problem is framed as a lack of individual competence, more user training becomes the answer. If the risks are described as technical deviations that can be managed through companies’ own safety systems, the larger questions of power, dependency, transparency and democratic oversight remain outside the frame.

In the essay Who Gets to Build the Future?, I explore this argument in greater depth. The point is not to distrust all technological development or to assume that companies’ concerns are insincere. Obama himself emphasises that he believes the concerns expressed by leading AI companies are genuine. But genuine concern does not alter the fundamental distribution of responsibility. Companies can contribute expertise about the technology and its risks, but they cannot be allowed to define on their own what kind of society that technology should help create. That is a democratic task, and to fulfil it, society’s institutions need knowledge, the capacity to act and the ability to respond to developments moving considerably faster than our conventional processes for governance and legislation.

Society’s Capacity Needs to Catch Up

The rapid growth in AI use is therefore neither unequivocally good nor bad news. It is a sign that the technology has already become part of everyday life and that the time for treating AI as a separate issue belonging to the future is over. We now need to build the shared capacity that ensures its use actually creates value. That requires organisations capable of learning while development is underway, leadership teams that understand digitalisation as a matter of organisational governance rather than simply technology procurement, shared structures that prevent every employee and public authority from having to solve the same problems independently, and public policy willing to assess digitalisation by its consequences rather than simply by the extent of its adoption.

It also requires us to understand caution in the public sector correctly. The fact that public organisations cannot experiment with people’s data, rights and access to public services in the same way that a private company tests a new consumer product is not evidence of hostility towards technology. It reflects a different kind of responsibility. But if caution is not combined with investment in expertise, shared infrastructure and practical capability, it risks turning into passivity, while development continues regardless and public organisations become dependent on solutions shaped elsewhere.

The public task, then, cannot be reduced to either slowing development down or accelerating it. It is about setting direction, building institutional capacity and ensuring that technology is used in ways that strengthen society’s capabilities rather than gradually eroding them. Sometimes that requires regulation and restrictions, sometimes investment and shared solutions, and sometimes the courage to refrain from doing something that is possible but not desirable. Above all, it requires the ability to connect questions that have been treated separately for far too long: innovation and security, efficiency and democratic values, technological development and human needs.

Because when adoption outpaces society’s capacity, a vacuum emerges, and that vacuum is not filled by neutrality. It is filled by the actors who already own the infrastructure, by the business models already in place and by individuals trying, on their own, to understand and manage the consequences.

The question, then, is no longer whether AI will change society. It already is, right in front of us and at a pace that means yesterday’s questions about the future have already become part of everyday life. The question is whether our institutions, organisations and leaders can develop the capacity required to steer that change towards the society we actually want, before its direction becomes so deeply embedded in our systems and habits that it begins to seem inevitable.