AI Adoption: Economic Benefits vs Social Risks
The Short-Term Economic Upside
Few technologies have delivered corporate returns as rapidly as artificial intelligence.
Enterprises deploying generative AI report average returns of roughly 2.8x on investment, with payback periods often around 15 months. Adoption has followed accordingly. Nearly nine in ten companies now report using AI in at least one business function, signalling that the technology has moved decisively from experimentation to operational deployment.
The macroeconomic effects are already visible. Technology and AI-related capital expenditures contributed about 60 basis points to US GDP growth in the first half of 2025.
Operational productivity gains are equally striking. Advanced models can reduce task completion times by up to 80 percent in many knowledge workflows, while some technical environments report productivity improvements approaching tenfold gains.
If these improvements diffuse widely, labour productivity in advanced economies could rise by around 1.8 percent annually, roughly double recent historical rates.
In the short run, AI looks less like a disruption than a powerful economic accelerant.
The Emerging Long-Term Social Costs
The same technologies driving this productivity boom are simultaneously reshaping labour markets.
Estimates suggest 100 to 150 million jobs globally could face high exposure to AI automation within the next five years. Early signals are already emerging. Occupations most exposed to AI have experienced employment declines of roughly 13 percent, while employment among 22 to 25 year-olds in AI-exposed fields has fallen by about 16 percent since late 2022.
Historically, new technologies eventually created more jobs than they destroyed. Artificial intelligence may prove different for two reasons.
First, the speed of diffusion.
Second, the automation of cognitive work rather than purely physical labour.
Together these forces raise the possibility that labour markets may struggle to adapt as quickly as the technology evolves.
The Rise of the K-Shaped AI Economy
Perhaps the most striking effect of AI adoption is the emergence of a K-shaped distribution of outcomes.
Workers and firms able to integrate AI effectively – capital owners, highly skilled professionals and organizations with deep data infrastructure – are experiencing dramatic productivity gains.
Others face increasing pressure as routine cognitive tasks become automated.
The result is what some analysts describe as a “K-shaped economy on steroids”, where technological progress lifts one segment of society while placing downward pressure on another.
The divergence is also geopolitical. Countries with advanced digital infrastructure and AI capabilities are pulling further ahead of those without them, raising the prospect that artificial intelligence could widen global economic inequality.
The Governance Gap
Beyond labour markets, AI introduces another and less visible risk: the erosion of institutional guardrails.
The rapid spread of generative models has sharply reduced the cost of producing sophisticated misinformation, synthetic media and automated fraud. Between 2022 and 2025, reported AI-related incidents and hazards increased by roughly 340 percent.
Technology companies are already responding with large defensive investments. Major platforms now spend $12 to $15 billion annually on trust and safety operations, roughly half of which is devoted to combating manipulation and disinformation.
For corporations, managing these risks increasingly resembles a structural cost of deploying AI. In heavily regulated sectors, the emerging “AI governance tax” can absorb between 8 and 12 percent of corporate revenue, while enterprise compliance tools can cost $500,000 to $800,000 annually.
Even the research ecosystem reflects this imbalance. Recent analysis by AIMGÂ examining 9,439 generative AI research papers published between 2020 and early 2025 finds that only 1,178 focus on safety, risk or governance, while 8,261 concentrate on capabilities or applications. In effect, the research community is producing roughly seven papers on advancing AI for every one examining its risks. Safety-related work accounts for just 12.5% of total output in the dataset. The figures do not represent a census of global research, but they suggest that far more intellectual capital is being invested in accelerating AI than in understanding how to govern it.
Lessons from Previous Technology Cycles
History suggests that regulation tends to lag innovation.
Automotive safety measures took nearly a decade to become mandatory after their invention. Tobacco regulation lagged scientific evidence for decades, eventually imposing enormous healthcare costs. Social media spread globally long before governments understood its societal consequences.
Artificial intelligence may now be following a similar trajectory.
Technological capability is advancing rapidly. Institutional adaptation is moving far more slowly.
The Strategic Question
Which brings us back to the central issue.
Are we approaching the point where the long-term social risks of AI begin to outweigh its short-term economic gains?
For now, the global economy still benefits from the productivity boom. Investment is rising, corporate margins are improving and innovation continues to accelerate.
But beneath this prosperity, structural tensions are building.
Labour displacement, widening inequality, geopolitical divergence and information instability are all accumulating beneath the surface of the AI boom.
If governments modernize regulation, corporations invest in responsible deployment and societies accelerate reskilling, artificial intelligence could unlock one of the largest productivity expansions in modern history.
If they do not, the current boom may prove to be only the opening phase of a far more disruptive transition.
Source: AIMG Research