AI: Exponential hype meets a linear reality?

Several Danish business leaders have recently warned that AI represents an exponential revolution that is approaching fast and could lead to fundamental changes, potentially making many jobs obsolete and even posing risks to humanity.

In reality, however, the spread of AI is progressing slowly. Much more slowly than expected, and the documented productivity gains remain limited. A recent analysis by The Economist shows that productivity growth in the United States remains broadly in line with the long-term average since 1950.

This has led some observers to fear an AI bubble, or at least that we may already have passed the peak of inflated expectations. Technology optimists, however – particularly technology investors – see the situation differently. For them, the absence of visible productivity gains is a sign that the real impact is still ahead.

Their argument reflects a familiar tech-optimist belief: exponential developments often appear slow and almost invisible at first. But eventually they accelerate rapidly, at a pace that seems to exceed imagination. Even small increases can become very large when growth is exponential, because the speed increases alongside the scale, unlike the more predictable, linear growth we usually observe.

The challenge with this view is that truly exponential functions rarely exist in nature or in real life. Bacteria can grow exponentially until they run out of food. Viruses can spread exponentially until populations develop immunity. In practice, most growth patterns follow a sigmoid or S-shaped curve: an exponential phase followed by a plateau as the system approaches its carrying capacity.

If we accept that AI is currently in an exponential phase, the real discussion is therefore not how extraordinary or frightening AI will become in the near future. The real question is when AI will reach its limits and flatten into a more stable phase.

In many ways, we may already be approaching that point. The limits of the AI revolution are not technological – they are defined by how we use the technology. No matter how advanced a technology becomes, it has little impact if organisations do not use it.

This shifts attention to the real challenges surrounding AI.

That is also the picture emerging in Nordic organisations. In a recent analysis by PA Consulting, 53 per cent point to lack of transparency and predictability as the biggest barrier to using AI. 46 per cent highlight human factors such as culture, skills and resistance to change. Only around a third report genuine technical limitations, and those mainly relate to data.

In other words: even if AI technology continues to develop exponentially, adoption does not automatically follow. Growth is constrained by trust, governance, skills and implementation – not computing power.

The exponential phase does not end because algorithms slow down. It ends when organisations reach their practical capacity to absorb and use the technology.

This does not mean the exponential phase cannot be extended. It simply requires addressing the real challenges around implementing and using AI, rather than focusing on narratives of doom or salvation.

This article was first published in Version2 in Danish.