Artificial intelligence: a disruption that is slower than it seems
In recent months, the debate around artificial intelligence has tended to swing between two unhelpful extremes. On the one hand, the promise that it will change everything tomorrow; on the other, a narrative of fear warning that it will destroy the market. Both views fail for the same reason: they confuse availability with real-world adoption. AI is already a tangible and deeply disruptive reality, but its impact will not be instantaneous, nor will it unfold exactly as it is currently being portrayed.
This confusion between availability and implementation is not new. We have seen it before with autonomous driving, which was presented as imminent and yet remains partial, limited, and highly dependent on context. Artificial intelligence is far closer to that kind of long, incremental and non-linear evolution than to a switch that can simply be turned on overnight.
The four gateways to AI adoption
To understand how AI will affect business, it is worth organising the discussion. There are four factors – four ‘gateways’ – that must be opened for a project to move from being a proof of concept to having a real impact: technological maturity, risk and the regulatory framework, infrastructure, and organisational culture. When one of them remains closed, the use case is delayed. When two people come together, the project remains nothing more than a well-designed presentation.
AI only delivers real impact when four gates are opened: technology, risk, infrastructure, and culture
The first step is technology. There’s a clear catch here. Some of the capabilities currently attributed to AI are real, but others are still in development, and some still belong more to the realm of science fiction than to engineering. Furthermore, the sales pitch usually comes from the seller of the tool, which is simply the nature of the market.
That is why it is essential for company management to distinguish between what works in the laboratory and what is ready for production, with guarantees of stability, cost-effectiveness and scalability.
The second hurdle is risk. In critical environments, you cannot simply connect a ‘black box’ to a banking system, a hospital’s intensive care unit or an energy infrastructure and hope for the best. The problem isn't a technical one; it's a question of responsibility. The benefits of automation can never outweigh the potential cost of error. In regulated industries, this principle determines the actual timeline for adoption.
The third factor is probably the most decisive: infrastructure. There are currently not enough data centres, chips, computing power or energy to scale up an ‘AI for everything’ immediately. Building a data centre takes between two and three years in the best-case scenario. It takes around a decade to build a semiconductor factory. And energy networks and power stations even more so, in addition to the prior social consensus that needs to be established. AI is also all about kilowatts.
The benefits of automation with AI can never outweigh the potential cost of error.
The fourth gateway is cultural. This is where many projects fail. Not for lack of algorithms, but because of resistance to change. Digital transformation has already shown us that technology is usually the easy part, and that the real challenge lies in changing how decisions are made, how progress is monitored and how results are measured.
With artificial intelligence, the leap is even greater, because redefining decision-making and control processes is essentially human work.
Real impact: business and employment
Will AI impact employment? It already is, although for now mainly in more mature organisations and not with the same intensity everywhere. Certain tasks face clear risk, particularly those that are repetitive and low value-added—for example, basic testing, incident classification, or highly mechanical functions in contact centres. In these cases, automation is a logical outcome.
However, business does not disappear; it evolves. We can see this in software development: writing code is becoming partly automated, but the work is shifting towards clearly defining specifications and validating outcomes. At the same time, new roles are emerging—profiles that did not previously exist—focused on ensuring AI is effectively integrated into business processes or on redesigning those processes around it.
AI brings less repetitive work, and more work at the intersection of technology, business and risk
It is how each company navigates this transformation that will make all the difference.
The question is not whether AI will change the world, but how it will do so and at what pace. And in the business world, the answer will come from those who execute best, not those who make the most noise.
The winner will be whoever applies a clear approach to their strategy, with sustained investment, realistic timelines, and an idea that is far from grand but highly profitable: but not today.