There’s been much talk recently – especially among politicians – about productivity. And for good reason: Australia’s labour productivity growth sits at a 60-year low.
To address this, Prime Minister Anthony Albanese has convened a productivity round table next month. This will coincide with the release of an interim report from the Productivity Commission, which is looking at five pillars of reform. One of these is the role of data and digital technologies, including artificial intelligence (AI).
This will be music to the ears of the tech and business sectors, which have been enthusiastically promoting the productivity benefits of AI. In fact, the Business Council of Australia also said last month that AI is the single greatest opportunity in a generation to lift productivity.
But what do we really know about how AI impacts productivity?
What is productivity?
Put simply, productivity is how much output (goods and services) we can produce from a given amount of inputs (such as labour and raw materials). It matters because higher productivity typically translates to a higher standard of living. Productivity growth has accounted for 80% of Australia’s income growth over the past three decades.
Productivity can be thought of as individual, organisational or national.
Your individual productivity is how efficiently you manage your time and resources to complete tasks. How many emails can you respond to in an hour? How many products can you check for defects in a day?
Organisational productivity is how well an organisation achieves its goals. For example, in a research organisation, how many top-quality research papers are produced?
National productivity is the economic efficiency of a nation, often measured as gross domestic product per hour worked. It is effectively an aggregate of the other forms. But it’s notoriously difficult to track how changes in individual or organisational productivity translate into national GDP per hour worked.
AI and individual productivity
The nascent research examining the relationship between AI and individual productivity shows mixed results.
A 2025 real-world study of AI and productivity involved 776 experienced product professionals at US multinational company Procter & Gamble. The study showed that individuals randomly assigned to use AI performed as well as a team of two without. A similar study in 2023 with 750 consultants from Boston Consulting Group found tasks were 18% faster with generative AI.
A 2023 paper reported on an early generative AI system in a Fortune 500 software company used by 5,200 customer support agents. The system showed a 14% increase in the number of issues resolved per hour. For less experienced agents, productivity increased by 35%.
But AI doesn’t always increase individual productivity.
A survey of 2,500 professionals found generative AI actually increased workload for 77% of workers. Some 47% said they didn’t know how to unlock productivity benefits. The study points to barriers such as the need to verify and/or correct AI outputs, the need for AI upskilling, and unreasonable expectations about what AI can do.
A recent CSIRO study examined the daily use of Microsoft 365 Copilot by 300 employees of a government organisation. While the majority self-reported productivity benefits, a sizeable minority (30%) did not. Even those workers who reported productivity improvements expected greater productivity benefits than were delivered.
Lukas Coch/AAP
AI and organisational productivity
It’s difficult, if not impossible, to attribute changes in an organisation’s productivity to the introduction of AI. Businesses are sensitive to many social and organisational factors, any one of which could be the reason for a change in productivity.
Nevertheless, the Organisation for Economic Co-operation and Development (OECD) has estimated the productivity benefits of traditional AI – that is, machine learning applied for an industry-specific task – to be zero to 11% at the organisational level.
A 2024 summary paper cites independent studies showing increases in organisational productivity from AI in Germany, Italy and Taiwan.
In contrast, a 2022 analysis of 300,000 US firms didn’t find a significant correlation between AI adoption and productivity, but did for other technologies such as robotics and cloud computing. Likely explanations are that AI hasn’t yet had an effect on many firms, or simply that it’s too hard to disentangle the impact of AI given it’s never applied in isolation.
AI productivity increases can also sometimes be masked by additional human labour needed to train or operate AI systems. Take Amazon’s Just Walk Out technology for shops.
Publicly launched in 2018, it was intended to reduce labour as customer purchases would be fully automated. But it reportedly relied on hiring around 1,000 workers in India for quality control. Amazon has labelled these reports “erroneous”.
More generally, think about the unknown number (but likely millions) of people paid to label data for AI models.
John G. Mabanglo/EPA
AI and national productivity
The picture at a national level is even murkier.
Clearly, AI hasn’t yet impacted national productivity. It can be argued that technology developments take time to affect national productivity, as companies need to figure out how to use the technology and put the necessary infrastructure and skills in place.
However, this is not guaranteed. For example, while there is consensus that the internet led to productivity improvements, the effects of mobile phones and social media are more contested, and their impacts are more apparent in some industries (such as entertainment) than others.
Productivity isn’t just doing things faster
The common narrative around AI and productivity is that AI automates mundane tasks, making us faster at doing things and giving us more time for creative pursuits. This, however, is a naive view of how work happens.
Just because you can deal with your inbox more quickly doesn’t mean you’ll spend your afternoon on the beach. The more emails you fire off, the more you’ll receive back, and the never-ending cycle continues.
Faster isn’t always better. Sometimes, we need to slow down to be more productive. That’s when great ideas happen.
Imagine a world in which AI isn’t simply about speeding up tasks but proactively slows us down, to give us space to be more innovative, and more productive. That’s the real untapped opportunity with AI.
