Speaking to an audience of executives gathered on Thursday at the Horizon Stage at Nexus Luxembourg, Emmanuel Vivier made no attempt to convince his audience that artificial intelligence represents a major game-changer. For the co-founder of the Hub Institute, who has been supporting large companies in their digital transformation for 20 years, this debate is already a thing of the past.
The technological capabilities are there. Investment is following suit. The real challenge now lies elsewhere. “The scarce resource isn’t AI, it isn’t computing power, it isn’t energy. It’s courage,” he said.
The courage to look beyond technological demonstrations and marketing hype. “Buying a few licences for AI assistants or conversational models doesn’t transform a business. That’s the easy part,” he insisted. The challenge, therefore, lies in reviewing processes, rethinking working methods and transforming organisations. This is a slower, more complex and often more political undertaking, as behind the tools quickly emerge issues of skills, training, recruitment and, at times, the redeployment of staff.
To gauge the scale of the current momentum, Emmanuel Vivier pointed out that Meta, Alphabet, Amazon and Microsoft alone are planning to invest nearly $700bn by 2026. In relation to the US economy, the investment in AI infrastructure would represent 1.9% of GDP, compared with 0.4% for the Manhattan Project and 0.6% for the Apollo programme. Another indicator, he argues, illustrates the scale of the shift: this year, more data centre projects are said to have been launched than office projects.
The question of a potential bubble was not sidestepped, with companies such as SpaceX, OpenAI, Anthropic and xAI preparing for IPOs with spectacular valuations. But for the speaker, the issue is almost secondary. “Even if not everything succeeds, there is an acceleration within the acceleration.” This wave of investment is fuelling the entire ecosystem, from chip manufacturers to energy suppliers, right through to models that see their capabilities double roughly every six months.
From AI that converses to AI that acts
For Emmanuel Vivier, this is where the real turning point lies. “2025 was the year when AI could converse. 2026 is the year when AI can take action.” The arrival of agents capable not just of performing a one-off task, but of managing an entire process, connecting to a CRM or SAP system, reasoning and taking action using tools, is changing the nature of the game for businesses.
Certain sectors appear to be particularly vulnerable to this trend. “Finance, law, insurance: you are the perfect target,” he said. Precisely because these sectors are regulated, standardised and structured, and therefore perfectly comprehensible to AI.
However, the expert urged caution regarding promises of immediate gains. According to a study of the 40 largest French companies, the productivity gains observed currently range between 1% and 7% per year. These results may seem modest at first glance, but they can become significant over time. Provided, however, that the transformations are properly supported. “Never underestimate the effort required to drive change. Simply making tools available is not enough: training – and retraining – is also essential.”
Luxembourg: a trusted jurisdiction
It was undoubtedly when he mentioned Luxembourg that his message struck a chord with the audience.
Whilst some see European regulation as a hindrance compared to the United States or China, Emmanuel Vivier sees it as a potential competitive advantage. Mastery of the GDPR, implementation of the AI Act, sector-specific regulatory expertise, digital infrastructure, multilingualism and political stability are, in his view, all assets that could position the Grand Duchy as a gateway to the European market. “Do you want to be a victim of AI or take advantage of it?” he asked.
According to the expert, these regulatory advantages are complemented by Luxembourg’s ability to serve as a testing ground. He specifically cited Pony.ai’s autonomous vehicles, for which the country became one of the first open-environment test sites in Europe at the beginning of June. “You don’t have to be big to be innovative,” he summed up.
This optimistic view does not, however, prevent him from highlighting several limitations. AI relies on massive amounts of computing power, and therefore financial resources. As companies in the sector are required to demonstrate their profitability, usage costs are likely to continue rising, to the point where, in some cases, human labour will once again become competitive. “We are rediscovering that humans, sometimes, aren’t that expensive,” he observed.
It is precisely this role of the human factor, he argues, that raises an even more worrying risk: the threat to junior staff. Automation primarily affects routine tasks that traditionally serve as a gateway into many professions. “If we no longer train junior staff, how will we ever have senior staff?” he warned, arguing for the retention of human oversight throughout the decision-making process.
Despite these reservations, his conviction is summed up in a phrase he is fond of: ‘return on intelligence’ rather than return on investment. The technologies exist, the capital has been raised, and use cases are multiplying. The question remains whether leaders will be able to drive the necessary transformations… before the pace of change is forced upon them.
