Hardly a week goes by at Glion without someone asking the same question: is artificial intelligence coming for our jobs, or for our clients? Nobody can say for sure, and anyone who answers with great confidence is probably selling something. But both questions may miss the point. To understand what AI changes, we need to look at what it is, what it does well, and where it creates tensions for business.

Strip away the mystique and today’s AI systems are statistical models of remarkable power. They learn patterns from vast amounts of data and use them to make predictions: the next word in a sentence, a customer’s likely preference, or the chance that a booking will be cancelled. In other words, AI is extraordinarily good at learning from the average and depict what is typical. That single fact explains much of what it can and cannot do for business, and it runs through everything that follows.

This also explains why AI matters so much. Prediction is needed almost everywhere, which is why economists tend to place AI among the general-purpose technologies (GPTs): technologies such as electricity and the internet that spread across the economy, improve over time, and trigger waves of complementary innovation across sectors. History offers important lessons about GPTs, so looking at the past can help us anticipate how AI may affect industries in the future.

The first lesson: Reorganising value creation

The first lesson from GPTs is that their value rarely goes to whoever adopts them first. It goes to whoever uses them to reorganise how value is created and captured.

Hospitality learned this the hard way. When the internet arrived, many hotels treated it as just another sales channel. Online travel agencies saw something else: an opportunity to reorganise the market around travellers’ searches. Network effects did the rest, and by the time major hotel groups launched their own search platforms, the game was largely decided. Luxury fashion houses, by contrast, reorganised to protect their brands, built their own digital channels, and retained control of their relationships with clients.

Thus, with AI, the strategic question is the same: not “How do we use it?” but “How can a company reorganise around it to shape the industry?”, a question that can be answered from the internal perspective of how companies organise tasks and human capital, but also from the external perspective that involves customers.

Reorganisation starts inside the firm, and here the evidence is encouraging. Research by Brynjolfsson and colleagues shows that customer-service agents’ productivity increases by about 15% when they use AI, with the largest gains among the least experienced. That is exactly what a technology that learns the “average” should do: spread the better habits of more effective agents to their lower-performing colleagues. This helps lift the floor of productivity.

Raising the ceiling, where the best people make the difference, especially in luxury services, is nonetheless another matter. The reason is that AI’s abilities are uneven. Studies by Dell’Acqua and colleagues describe a “jagged frontier”: AI excels at some tasks but can mislead on others that appear just as easy. The challenge is knowing where a given task sits in relation to this frontier and, therefore, how well AI is likely to perform. This is non-trivial for three reasons. First, the frontier is invisible. Second, AI’s ability at solving tasks evolves with the models. Third, we tend to “fall asleep at the wheel”, passively accepting AI suggestions without critically assessing them.

Reorganising around AI is therefore neither replacing people nor simply giving them the technology. It is about knowing which tasks to delegate to AI, which to share with it, and which to keep with people. That judgement is fast becoming a core management skill.

The same logic applies, with even greater force, when we take the external perspective and turn from employees to customers. Here, AI’s great promise is personalisation: treating every customer as an individual, at scale.

Yet AI learns from data, and data describe the many rather than the one. The more a system relies on general patterns, the greater the risk that its answers tend towards what is typical, just when customers want to be recognised for who they are. The promise of personalisation built on the average therefore presents a tension that technology alone cannot resolve. This tension is particularly acute in luxury, where clients pay precisely not to be treated as average, yet may also be more sensitive to the use of personal data that can help AI tailor its responses to them.

Overall, research and practical use cases point to a clearer direction for reorganising internal work around AI. By contrast, companies are still experimenting cautiously with AI applications aimed at customers.

The second lesson: The power of complementary assets

The second lesson from GPTs is that, when a technology becomes widely available and cheap, the things that complement it become more valuable. For AI, these complementary assets are high-quality data and human managerial direction.

Acquiring and organising these complements takes time. When electricity reached factories, the main benefits arrived only decades later, after operations had been redesigned to take advantage of it. AI may be at a similar point. A July 2026 report by Bain & Company and the Comité Colbert found that around 60% of luxury groups and maisons had not yet seen a significant impact from any AI deployment. Companies have only scratched the surface of this reorganisation, and the innovations that will define the next decade may therefore not yet exist.

The overarching lesson from GPTs is that the value of a technology does not necessarily go to the first adopter, but to whoever uses it to reorganise how value is created and captured. What, then, does it take to reorganise around AI?

We must return to the idea of AI as a set of statistical models of remarkable power. A statistical model is only as good as the quality of the data on which it is trained. Moreover, it has no purpose in itself unless it is assigned an appropriate task. To reorganise effectively, companies need high-quality data, and managers must deliberately direct how AI is used, balancing human and AI contributions according to the task. Ultimately, this means directing AI to automate the “average” tasks, thereby freeing human professionals to deliver the bespoke, non-average touches that define exceptional service.

Recognising tasks and directing AI is exactly what we teach students and executives at Glion.

Alessio Delpero is Associate Professor at Glion Hospitality & Luxury Business Education.