The AI race that Europe is capable of winning
Leadership will depend not only on the power of the models, but on the value generated. The Specialist AI approach can transform Europe’s industrial heritage into a competitive advantage.
To lead the way in artificial intelligence, Europe must not try to outdo the United States and China by pouring billions into state-of-the-art models. It must focus on its own competitive advantages. Chasing AI investment without a clear strategy would mean playing the wrong game – at a high cost. There is another competition, in which what matters is not how much companies spend or how powerful the models they use are, but the value they manage to generate. The winner is whoever can choose the most suitable artificial intelligence for each task, offering the best balance between results and resources. It is a game in which Europe has what it takes to succeed.
Centuries of industry, research, precision engineering and craftsmanship have endowed Europe with a heritage that is difficult to replicate: proprietary knowledge, high-quality data and specialist expertise. Businesses and institutions hold projects, engineering data and operational expertise that cannot be replicated simply by increasing computing power. And as general-purpose models become more capable and interchangeable across many tasks, this heritage will become even more valuable.
This is where the concept of Specialist AI comes into play: artificial intelligence built around the data, knowledge, processes and specific constraints of a sector or organisation, combining the models and infrastructure best suited to the tasks at hand. Its strength lies not in the scale of the models, but in its ability to improve processes, boost productivity, accelerate innovation and make use of knowledge that is often scattered throughout organisations.
The European Patent Office already processes 400,000 pages a day using a model designed for a specific purpose, with a billion parameters and an error rate of less than 1 per cent. This proves that specialisation does not mean compromising on results.
The idea that bigger is always better does not hold water. Yet many organisations direct even simple requests – from drafting emails to analysing documents – towards expensive state-of-the-art models. Our research shows that one in three organisations exhausts its token budget before the end of the year, whilst less than 20 per cent of expenditure can be linked to a measurable financial return. Another of our analyses, covering over 9,000 tasks, found that around 90% did not require the advanced capabilities of state-of-the-art models.
