In the following analysis, Raphael Olszyna-Marzys, international economist at J. Safra Sarasin Sustainable Asset Management assesses where we are with AI and whether it really can deliver a boost to growth
Since the mid-2000s, the global economy has endured the weakest productivity growth in decades. Long-term structural trends such as geo-economic fragmentation and ageing populations are likely to further dampen growth and exacerbate inflationary pressures. Artificial intelligence (AI), by contrast, is a source of optimism. It holds the potential to accelerate productivity growth and reduce costs.
AI, as ChatGPT explains, is ‘the simulation of human intelligence by machines, to perform tasks such as learning, reasoning, problem-solving, perception, and language understanding.’ This broad field encompasses technologies like machine learning, computer vision, and deep learning, many of which have been in use for years. Recent advances in large language models and chatbots, exemplified by ChatGPT, have drawn attention to Generative AI (GenAI). Although it represents only a subset of current AI applications, it has fuelled hopes for a productivity renaissance and spurred a surge in AI-related investment, particularly in the United States over the past two years.
As a general-purpose technology, akin to electricity or the internet, AI has the potential to boost productivity through two main avenues. The first is by enhancing the efficiency of existing jobs and tasks, with some roles becoming fully automated. Automation is not new, but AI could extend it to cognitive tasks once deemed beyond its reach. For instance, studies suggest that GenAI can improve productivity in activities like coding and writing by 25%. The second avenue, which excites economists the most, lies in AI’s capacity to generate new ideas and innovations. Total factor productivity (TFP)—the growth in output not attributable to additional capital or labour—has historically alternated between strong and weak phases. By reducing the cost of innovation and overcoming barriers to technological progress, AI could push the global economy onto a much higher growth trajectory.
Quantifying AI’s impact on overall productivity remains challenging, even when focusing solely on its ability to enhance existing tasks. The outcome depends not only on how widely the technology is adopted but also on how intensively it is used. So, what can be said about these two critical factors?
Recent surveys indicate a swift rise in the use of GenAI, largely driven by free tools adopted informally at workplaces and homes, rather than by paid-for business applications. In a recent paper, Bick, Blanding, and Deming of the St. Louis Fed reveal that by August 2024, nearly 40% of Americans aged 18 to 64 had used generative AI to some extent, with about a third reporting daily or frequent weekly usage. Tasks ranged from drafting communications to searching for information and managing administrative work. Notably, at least one in four respondents used AI for each task presented.
In contrast, businesses have been slower to integrate AI into their operations. According to the Business Trends and Outlook Survey, updated in early December, only 5% of US firms use AI in production, with 6.5% planning to do so within six months. Uncertainty about its future use remains high, with 23% of companies still unsure whether they will use AI in the near future. This cautious approach reflects a mix of growing awareness of AI’s potential to enhance profitability and uncertainty over its best applications. A Morgan Stanley survey of 400 companies found that strategy gaps, lack of management buy-in, and data security concerns are key barriers to adoption. For those that have embraced AI, outcomes have been promising: 50% of projects met expectations, and 40% exceeded them. Revenue and cost considerations vary by company size. Larger firms focus on cutting costs with AI, while smaller ones balance cost-saving and revenue-generation goals. Technology and industrial sectors lead in pursuing revenue gains. A McKinsey study based on 2022 data highlights AI’s impact on profitability, with 42% of companies reporting lower costs and 59% noting higher revenues.
Outside of the US, GenAI adoption has spread most rapidly in China. According to a survey of large companies around the world by SAS and Coleman Parkes Research, an AI and analytics company, 83% of Chinese companies in the survey have begun to implement GenAI, compared to 65% in the US. While China’s development of large language models still lags behind the US, it excels in embracing AI applications for specific use. This is in line with the Chinese government’s objective of leveraging AI to transform existing industries. The AI plus initiative, introduced in the Government Work Report in March this year targets the integration of AI and the real economy. Industrial upgrading and modernising to ‘smart manufacturing’ to improve production efficiency and product quality is the main goal. Companies in the real economy (such as electric vehicle producers) have collaborated with AI developers (such as Baidu) to launch AI-driven products like robotaxis.
A good example of China’s integration of AI into its industries is its rapid implementation of smart industrial robots, with China leading the rest of the world in this area. Clearly the strategy also serves its goal to remain the factory of the world even as its population ages and working cohorts decline. The latest humanoid robots, driven by AI, aim to improve precision in tasks such as product inspection.
GenAI may already be contributing to recent productivity growth in the US. Productivity has risen at an annualised rate of 2% over the past two years—double the average of the previous two decades. Bick, Blanding, and Deming estimate that GenAI is aiding 0.5–3.5% of all working hours. Combining this with some of the boost in task productivity identified in various surveys suggests that AI’s current use is already boosting labour productivity by 0.1-0.9%. If adoption rates continue to rise quickly, a period of robust productivity gains seems possible.
A paradox underlies GenAI’s ascent. Individual users, aided by free tools, have driven adoption, while businesses lag in implementing paid AI solutions. This dynamic raises questions about the return on investment for tech giants—Amazon, Microsoft, Alphabet, and Meta—that have poured vast sums into AI development. These firms now account for over 15% of capital expenditures by US nonfinancial companies, yet their returns remain uncertain. Moreover, training large language models is increasingly costly and constrained by infrastructure limits.
The risk is that adoption may not scale fast enough for AI providers to recoup their investments, offset depreciation, and sustain margins. This could lead to slower investment or aggressive monetisation efforts, delaying broader adoption and productivity gains. The optimists envision a different trajectory. GenAI applications, such as autonomous agents and robotaxis, could soon deliver strong cash flows, enabling continued investment in the sector. If realised, this scenario would support the ongoing rapid adoption of AI and bolster productivity growth. In short, the coming years will test whether businesses can match individual users’ enthusiasm for AI and whether tech giants can convert their vast capital outlays into durable returns. In either case, AI is poised to remain a pivotal force driving technological progresses and the global economy.
