Generative artificial intelligence promises to lift global productivity, but the gains will materialise only if companies and governments can redeploy workers effectively rather than simply automate tasks, according to a report on the technology’s economic impact by McKinsey & Company.
The adoption of generative AI in the workplace has accelerated at a pace rarely seen for new technologies. By early 2024, about 75% of knowledge workers globally reported using generative AI tools at work, nearly double the share recorded just six months earlier.
The speed shows the scale of the shift. ChatGPT reached one million users in five days. AI tools are embedded in daily work.
According to a publication by Upwork, corporate leaders have moved just as fast. Under pressure to lift efficiency in a slowing global economy, companies are turning to AI to raise output without expanding headcount. Almost all executives surveyed, about 96%, say they expect generative AI to boost productivity, signalling broad confidence that the technology will reshape how work gets done.
McKinsey & Company’s study finds that generative AI can automate or augment a large share of work activities across the global economy, potentially adding meaningful growth at a time when ageing populations and slowing labour-force expansion are weighing on output. But it cautions that productivity gains are not automatic. They depend on whether labour hours freed by automation are shifted into tasks that are at least as productive as those they replace.
Without that transition, AI risks becoming a cost-saving tool that delivers limited macroeconomic benefit, while increasing disruption in labour markets.
The report estimates that automation powered by generative AI and related technologies could add between 0.5% and 3.4% to annual global productivity growth through 2040, depending on how quickly adoption spreads. Generative AI itself accounts for a smaller but still significant share of that uplift. Crucially, those gains assume workers can move into new or redesigned roles without a loss in productivity.
That assumption is doing much of the work.
Historically, productivity shocks from new technologies have taken years to feed through to growth as firms restructured processes and workers acquired new skills. The report suggests generative AI may compress those timelines, but also raises the risk that labour markets struggle to adjust at the same speed as technology adoption.
For businesses, the findings challenge the idea that deploying AI tools alone will deliver higher output. Productivity gains require complementary investments in training, job redesign and organisational change. Firms that automate tasks without redefining roles may reduce headcount or hours worked, but fail to increase output per worker.
The implications are particularly acute for white-collar sectors. Unlike earlier waves of automation that targeted routine manual or clerical work, generative AI affects higher-paid cognitive tasks in areas such as finance, marketing, customer operations and software development. In many cases, workers remain in the same occupation but see their mix of activities change substantially.
That creates both opportunity and risk. Workers who can use AI effectively become more productive and potentially more valuable. Others may find parts of their roles hollowed out, forcing transitions to new functions or occupations. Managing that shift will determine whether AI lifts wages and growth or deepens inequality.
For policymakers, the message is that AI adoption is not a substitute for labour-market reform. Education systems, retraining programmes and mobility support become central to capturing productivity gains. Without them, economies could see higher displacement with little improvement in growth.
The report also highlights a divergence between countries. Economies with strong institutions, flexible labour markets and established training systems are better positioned to redeploy workers into higher-value tasks. Others risk seeing automation translate into job losses or underemployment rather than higher output.
In that sense, generative AI is as much a test of management and policy capacity as it is of technological capability. The productivity dividend is available, the report suggests, but it is conditional. Technology can free up time. Turning that time into growth remains a human decision.
