⚡ Digitalization is Dead. Long Live Automation.
The debate about what AI does to work, employment and social cohesion has been developed in two thesis papers in recent weeks. What intentionally remained in the background was the other side of the equation: What emerges? Who wins? And why does Europe have better cards in this game than the public debate suggests?
The answer begins with an observation that sounds like a provocation but is simply precise: Digitalization is, as a goal and as a promise, done. What is digitalized today is automatable tomorrow. What is not yet digitalized will become so on the way to automation. This is not a retrospective. It is the description of a threshold that is currently being crossed. What this means for Europe, for small businesses, for the software industry, and for the question of where the next competitive advantages will emerge, I would like to develop in six theses.
Thesis 1: Digitalization was the preliminary stage. Automation is the goal.
For decades, digitalization was considered the strategic goal. Mapping processes in software, replacing paper with data, converting analog workflows into digital ones. It was tedious, expensive, and took longer than planned in many companies. It was still the right thing to do. Not because digitalization was the goal, but because it created the prerequisite for what is coming now.
Because what is digitalized can be automated. Not someday, not as a future vision, but today, with available tools, at costs that were unthinkable two years ago. The difference from the earlier automation wave is not gradual, it is categorical. Earlier automation required specialized systems for narrowly defined tasks. What is now available are generalized systems that understand context, interpret tasks, and execute processes independently, without every step having to be manually coded beforehand.
This also means: Companies that have delayed their digitalization are not better off than thought. They have skipped the preliminary stage and land directly behind. And companies that are well digitalized are sitting on an asset whose value they have not yet fully recognized: structured, proprietary data that is exactly the raw material that AI systems need to become domain-specific and truly useful.
What is digitalized is automatable. What is not yet digitalized will become so on the way there.
Thesis 2: Europe does not need to hide. It has overlooked the wrong assets.
The public debate about Europe's position in the AI age revolves almost exclusively around what is missing: the large infrastructure platforms, the computing capacities, the venture capitalists willing to invest billions in early stages. This is not wrong. But it is only half the truth.
What Europe has is systematically underestimated in this debate. First: proprietary corporate data. In manufacturing, logistics, healthcare, and finance, European companies have built decades of structured, high-quality datasets that no American or Chinese model can replicate from the outside. This data is the real capital of the AI era, not the infrastructure that processes it. Second: engineering depth. Germany, Switzerland, the Netherlands, Scandinavia, Austria have been producing world-class engineers and software developers for decades. This quality is not self-evident. It is a competitive advantage that does not disappear in a world where software development is being democratized, but expresses itself anew.
The decisive point: Europe does not need Silicon Valley. It needs companies and individuals who understand what assets they already have and know the tools to deploy them now.
Europe's strength does not lie in the infrastructure it lacks. It lies in the data and engineering know-how it already possesses.
Thesis 3: Agentic AI democratizes software development. A small team today beats an entire department.
What has changed in software development in recent months can hardly be described in the usual language of technological progress. Projects that previously required teams of ten developers for six months are now realized by two people with agentic tools in weeks, sometimes days. This is neither exaggeration nor isolated case. It is the new normal for anyone who actually uses these tools.
The term that has established itself is Vibe Coding: the practice of developing software architectures and entire applications through natural language instructions to AI agents that write, test, correct, and iterate code. What is qualitatively new: the developer is no longer primarily a coder. They are a conductor. They define direction, evaluate results, ask the right questions of systems that handle execution. This shift fundamentally changes who can competitively develop software.
A single expert with deep domain knowledge and the right tools can replace departments today, not because they work harder, but because they operate a multiplier that did not previously exist.
The developer is no longer primarily a coder. They are a conductor. And a good conductor no longer needs a large orchestra.
Thesis 4: The mass extinction of IT consultancies has begun. It is no longer a forecast.
What applies to accountants and clerks applies to IT consultancies and system houses in an intensified form. The business model of this industry is based on a simple principle: selling complexity as a service. Implementations take a long time because they are difficult. Teams are large because the tasks require it. Hourly rates are high because expertise is rare.
All three premises are eroding simultaneously. Agentic AI dramatically compresses implementation times. The required team size decreases proportionally. And the expertise that previously required years of specialization becomes accessible through models that make domain-specific knowledge available on demand.
For European IT consultancies, there is an additional complication: workforce reduction at the pace the market change would require is legally and culturally difficult. This slows adaptation but does not solve the problem. Companies that do not radically rethink now will not slowly shrink. They will quickly become obsolete.
Thesis 5: Nearshoring and offshoring are structurally dead.
The promise of nearshoring and offshoring was always the same: quality at lower costs by moving work to where salaries are lower. This model had its justification as long as software development was primarily working time that could be geographically relocated.
This foundation no longer exists. When a developer in Hamburg achieves the productivity of ten developers with agentic tools, the cost advantage of a team in Warsaw, Bangalore, or Kyiv is structurally neutralized. Not because quality there has declined, but because the multiplier works equally everywhere.
Offshoring was an arbitrage model based on labor costs. This basis no longer exists once productivity multipliers are geographically equally distributed.
Thesis 6: A European software renaissance is imminent. Not despite missing infrastructure, but because of new tools.
At a certain point in the development of platform technologies, value creation shifts from infrastructure to application. The internet created Google and Amazon, not the companies that laid fiber optic cables. The smartphone created WhatsApp and Spotify, not the chip manufacturers. AI infrastructure will follow a similar trajectory. The question is not who owns the computing capacity, but who has the domain knowledge to build something valuable from it.
Europe has this domain knowledge. In industry, healthcare, finance, logistics. What was previously missing was access to development capacities that would have enabled small and medium-sized companies to convert this knowledge into scalable software. This access now exists. Anyone with an internet connection and the willingness to seriously learn agentic tools can now realize software projects that two years ago would have required a funded startup with a team of developers.
There is a further shift that is hardly discussed in public debate: apps and applications as we know them will come under pressure as a concept. The fixed interface that you open, navigate, and operate is a product of a world where software was built statically. What is emerging is something different: interfaces that arise situationally, built from data sources and services connected live via APIs, MCPs and comparable protocols.
Those who have domain knowledge and master the tools need neither Silicon Valley nor a large team. They need a good question and a good agent.
The shift that is taking place is not one where Europe has to watch. It is one where Europe can participate, if it stops measuring itself against what it does not have and starts working with what it already possesses. This does not require billion-dollar industrial policy. It requires individuals and companies who understand what is currently available and the willingness to seriously deploy it. The window is open. When it closes is unknown. That it will close is certain.
Sources
- McKinsey Global Institute: Agents, Robots and Us. November 2025.
- Andreessen Horowitz (a16z): The New Business of AI. 2025.
- Stanford HAI: AI Index Report 2025.
- GitHub / Microsoft: The State of the Octoverse 2025.
- European Commission: European AI Strategy and Data Act. 2024.
- Bitkom Research: KI-Nutzung in deutschen Unternehmen 2025.
- OECD: The Impact of AI on Productivity and Growth. 2025.
- Goldman Sachs Global Investment Research: AI and the Future of European Software. 2025.