Artificial intelligence could launch a new era of growth – but only if our institutions prove just as dynamic, says INSEAD Professor and Nobel laureate in Economics Philippe Aghion. That warning anchors “AI: Our Ambition for France”, the 2024 national report produced by an AI commission Aghion co-chaired for French President Emmanuel Macron.

The report argues that productivity gains from AI will depend less on algorithms and more on the institutions built around them. It also places AI alongside the great general-purpose technologies of history -- electricity, IT and the internet – that reshaped productivity, employment and innovation.

“AI is a general-purpose technology that affects all sectors of activity,” Aghion said in presenting the commission report in an INSEAD Tech Talk X earlier this year, months before he was announced as one of three joint winners of the Nobel Prize in Economics.

The speed at which AI, powered by generative models that can create text and images with striking realism, has been adopted is breathtaking. Netflix took two and a half years to reach one million users, Instagram two and a half months, and ChatGPT only five days.

Such speed, Aghion said, hints at AI’s power to reshape economies within a single decade.

How AI boosts productivity

According to Aghion, AI drives productivity in two ways. The first is familiar, as it involves the automation of tasks in producing goods and services. The second, newer and potentially more transformative, consists of the automation of tasks in the production of ideas.

Working with economists Ben Jones and Chad Jones, Aghion modelled how AI enhances both domains. Real-world studies have begun to confirm their expectations. Brynjolfsson et al. found that at a Fortune 500 company advising small businesses on enterprise software, employees given access to ChatGPT solved 14 percent more problems per hour after one month and a further 25 percent after two months. 

A similar pattern is emerging in research settings. One R&D lab testing AI tools for materials discovery reported that access to AI raised the number of new materials identified by 44 percent and number of patent filings by 39 percent.

Early adopters stand to gain most

For business leaders, the message is clear: Guided, early use of AI tools can deliver measurable gains long before the technology matures.

Extrapolating from earlier technological revolutions, Aghion estimated that AI could lift annual productivity growth by about 0.7 percentage points for a decade – a boost similar to that delivered by IT (0.8) or electricity (1.3). For France, that could mean an extra €250-€400 billion in national income after ten years.

But Aghion stressed that these gains will materialise only if the benefits are broadly shared and not captured by a few dominant players.

Why jobs may grow, not vanish

Technology anxiety is hardly new. The steam engine, electricity and industrial robots all provoked fears of mass unemployment. Yet in every case, new technologies created more jobs than they destroyed.

Drawing on data from French firms, Aghion found that companies using automation or robots often hired more people, as productivity gains made them more competitive internationally. A similar pattern now appears for AI.

Surveying 9,000 firms between 2018 and 2020, his team compared adopters of AI tools with matched non-adopters. Overall, employment rose in AI-using firms. Gains were strongest in technical and managerial roles and weaker in administrative and sales functions. But even occupations often labelled “at risk”, such as accountants and telemarketers, showed net positive demand once productivity effects were included.

Fewer than 5 percent of jobs, the report concludes, face genuine displacement risk. Most will be reshaped rather than eliminated. The key, again, is policy: With the right training and transition support, the benefits of AI outweigh its disruptive effects.

Denmark’s “flexicurity” model shows how this can work. Workers who lose their jobs receive up to 90 percent of their pay for two years while retraining. INSEAD’s Alexandra Roulet has demonstrated that such a safety net eliminates the health and death hazards that come with business shutdowns. 

Aghion contrasted this with the United States, where weaker safety nets have coincided with rising “deaths of despair”. Institutional design, not AI, determines whether disruption brings resilience or distress.

The lesson from the IT revolution

The 1990s IT boom began with soaring productivity but ended with rising market concentration. Quick to harness digital scale, firms like Google, Amazon, Microsoft and Walmart grew through mergers and acquisitions. Their dominance gradually discouraged new entrants.

Aghion showed that the rate at which new American firms were created fell sharply after 2000, while the average mark-up across firms rose. It was not that every company became more profitable; rather, high-mark-up “superstar” firms gained market share. The result was a slowdown in productivity growth even as digital technologies advanced.

Artificial intelligence, he warned, could follow the same path. A few global incumbents are already dominating upstream segments of the AI value chain, such as cloud computing and specialised chips. Without deliberate competition policy, the next decade could see another burst of growth followed by stagnation as concentration sets in.

A new policy frontier

To prevent that outcome, Aghion’s commission recommends a strategy that blends competition policy with industrial policy.

• Re-tool competition policy

Encourage open-source models and ensure access to training data. Extend the EU’s Digital Markets Act across the AI value chain, from cloud services to foundation models. Regulators must prevent incumbents from using data control to block new entrants. But they must also guard against overregulation, which hinders start-ups more than it restrains large firms.

• Invest in shared infrastructure

Computing power is now essential infrastructure. Governments should join forces with industry to build more local data centres, keeping control of key technologies while cutting AI’s energy use. The report calls for major investment in high-performance computing that is open to researchers and smaller firms alike.

• Strengthen human capital

From early schooling to lifelong learning, AI literacy must become universal. The report proposes an “AI exception” in public research funding so that top scientists can divide their time between universities and the private sector without losing academic standing. This hybrid model, common in the US, could help Europe retain its talent rather than lose it to Silicon Valley.

Together, these measures aim to foster innovation without creating monopolies – an equilibrium that eluded the IT era.

The real constraint

Asked whether AI’s promise will fade as ideas become harder to find, Aghion offered a measured reply. In his view, AI makes ideas easier to find by accelerating how knowledge is combined and reused.

What worried him is not technological exhaustion but institutional inertia – the risk that governments, regulators and education systems adapt too slowly. Growth, he argued, is always the joint result of technology and institutions. AI’s frontier is moving fast, and whether Europe keeps pace will depend on how quickly its institutional architecture evolves.

“The technology is revolutionary,” he concluded. “The question is whether our institutions will be able to reform fast enough to fully exploit it.”