The AI paradox: everyone uses it but few really exploit it

"We have noticed that generative AI can enable these logics even in contexts where it was previously difficult," Buganza explains, "within companies, it can help break down organisational silos by fostering collaboration between different functions.

A virtuous example of the combination of platform thinking and AI is that of Fujitsu. The Japanese company developed an internal platform capable of linking product use cases with the needs of salespeople. This solution, the result of a participative process in which employees experimented with customised agents, facilitated knowledge sharing and improved the organisation.

"This use is very simple from a technological point of view, but it is far from the predominant use that emerges from our research," says Trabucchi.

Another revelation of the study concerns the use of this technology in teams. According to the experience of the Platform Thinking HUB 2025, working in teams helps to consider AI as an additional member of staff, improving interaction with the model and fostering a more conscious and strategic use.

However, research shows that the adoption of this method is still limited. The majority of employees use AI alone for individual tasks. Here a problem emerges called the shadow AI effect, i.e. the use of personal tools instead of company tools. According to research by MIT, only 40 per cent of companies have professional subscriptions, while 90 per cent of employees secretly use their own tools.

"Personal technology is often simpler and more intuitive than corporate technology, so people continue to use external tools for work as well," says Buganza. "But this is a huge risk: it means putting sensitive data into public systems. A striking case is that of an American pharmaceutical company that uploaded confidential documents to Gemini and found them indexed on Google.

According to Daniel Trabucchi, careful training is needed to avoid this danger: 'The key word is awareness. Companies must explain to employees what can and cannot be done, and at the same time offer experimental and user-friendly tools. A combination of proprietary solutions and open tools with guarantees can work well'.

To get the most out of this technology, we need to move from a 'co-pilot' approach, which uses AI as an assistant to increase productivity, to 'co-thinking', which sees AI as a partner in the cognitive process.

"Many companies plan everything in advance and approach it in atop down way, with a few pilot projects and small working groups, but the world of AI changes every week. We need a more agile approach that allows employees to experiment: this is also a social technology that changes the way people work and think," Buganza concludes.

While the debate continues between those who announce the end of the hype for artificial intelligence and those who continue to support it, the Polytechnic's observatory continues to research the potential of AI, placing it in a broader framework of innovation through platform thinking. "When you introduce profound transformations, the results do not come immediately," the project directors emphasise.