Does AI actually increase productiveness? Thus far the science is displaying blended outcomes

There’s been a lot speak not too long ago – particularly amongst politicians – about productiveness.

And for good purpose: Australia’s labour productiveness development sits at a 60-year low.

To deal with this, Prime Minister Anthony Albanese has convened a productiveness spherical desk subsequent month. This may coincide with the discharge of an interim report from the Productiveness Fee, which is 5 pillars of reform. One in all these is the function of information and digital applied sciences, together with synthetic intelligence (AI).

This can be music to the ears of the tech and enterprise sectors, which have been enthusiastically selling the productiveness advantages of AI. The truth is, the Enterprise Council of Australia additionally mentioned final month that AI is the single biggest alternative in a era to carry productiveness.

However what do we actually learn about how AI impacts productiveness?

What’s productiveness?

Put merely, productiveness is how a lot output (items and companies) we will produce from a given quantity of inputs (resembling labour and uncooked supplies). It issues as a result of increased productiveness usually interprets to the next way of life. Productiveness development has accounted for 80% of Australia’s earnings development over the previous three a long time.

Productiveness will be considered particular person, organisational or nationwide.

Your particular person productiveness is how effectively you handle your time and sources to finish duties. What number of emails are you able to reply to in an hour? What number of merchandise are you able to examine for defects in a day?

Organisational productiveness is how properly an organisation achieves its objectives. For instance, in a analysis organisation, what number of top-quality analysis papers are produced?

Nationwide productiveness is the financial effectivity of a nation, typically measured as gross home product per hour labored. It’s successfully an mixture of the opposite types. Nevertheless it’s notoriously tough to trace how adjustments in particular person or organisational productiveness translate into nationwide GDP per hour labored.

AI and particular person productiveness

The nascent analysis inspecting the connection between AI and particular person productiveness reveals blended outcomes.

A 2025 real-world research of AI and productiveness concerned 776 skilled product professionals at US multinational firm Procter & Gamble. The research confirmed that people randomly assigned to make use of AI carried out in addition to a group of two with out. A related research in 2023 with 750 consultants from Boston Consulting Group discovered duties had been 18% sooner with generative AI.

A 2023 paper reported on an early generative AI system in a Fortune 500 software program firm utilized by 5,200 buyer help brokers. The system confirmed a 14% enhance within the variety of points resolved per hour. For much less skilled brokers, productiveness elevated by 35%.

However AI doesn’t at all times enhance particular person productiveness.

A survey of two,500 professionals discovered generative AI truly elevated workload for 77% of staff. Some 47% mentioned they didn’t know methods to unlock productiveness advantages. The research factors to boundaries resembling the necessity to confirm and/or appropriate AI outputs, the necessity for AI upskilling, and unreasonable expectations about what AI can do.

A latest CSIRO research examined the every day use of Microsoft 365 Copilot by 300 staff of a authorities organisation. Whereas the bulk self-reported productiveness advantages, a sizeable minority (30%) didn’t. Even these staff who reported productiveness enhancements anticipated better productiveness advantages than had been delivered.

AI and organisational productiveness

It’s tough, if not unimaginable, to attribute adjustments in an organisation’s productiveness to the introduction of AI. Companies are delicate to many social and organisational elements, any certainly one of which could possibly be the explanation for a change in productiveness.

However, the Organisation for Financial Co-operation and Improvement (OECD) has estimated the productiveness advantages of conventional AI – that’s, machine studying utilized for an industry-specific process – to be zero to 11% on the organisational stage.

A 2024 abstract paper cites impartial research displaying will increase in organisational productiveness from AI in Germany, Italy and Taiwan.

In distinction, a 2022 evaluation of 300,000 US companies didn’t discover a vital correlation between AI adoption and productiveness, however did for different applied sciences resembling robotics and cloud computing. Probably explanations are that AI hasn’t but had an impact on many companies, or just that it’s too laborious to disentangle the affect of AI given it’s by no means utilized in isolation.

AI productiveness will increase may typically be masked by further human labour wanted to coach or function AI techniques. Take Amazon’s Simply Stroll Out know-how for retailers.

Publicly launched in 2018, it was meant to cut back labour as buyer purchases can be totally automated. Nevertheless it reportedly relied on hiring round 1,000 staff in India for high quality management. Amazon has labelled these experiences “misguided”.

Extra usually, take into consideration the unknown quantity (however seemingly tens of millions) of folks paid to label knowledge for AI fashions.

AI and nationwide productiveness

The image at a nationwide stage is even murkier.

Clearly, AI hasn’t but impacted nationwide productiveness. It may be argued that know-how developments take time to have an effect on nationwide productiveness, as corporations want to determine methods to use the know-how and put the mandatory infrastructure and expertise in place.

Nonetheless, this isn’t assured. For instance, whereas there may be consensus that the web led to productiveness enhancements, the consequences of cellphones and social media are extra contested, and their impacts are extra obvious in some industries (resembling leisure) than others.

Productiveness isn’t simply doing issues sooner

The frequent narrative round AI and productiveness is that AI automates mundane duties, making us sooner at doing issues and giving us extra time for artistic pursuits. This, nevertheless, is a naive view of how work occurs.

Simply because you may take care of your inbox extra rapidly doesn’t imply you’ll spend your afternoon on the seaside. The extra emails you hearth off, the extra you’ll obtain again, and the unending cycle continues.

Quicker isn’t at all times higher. Typically, we have to decelerate to be extra productive. That’s when nice concepts occur.

Think about a world through which AI isn’t merely about rushing up duties however proactively slows us down, to present us area to be extra revolutionary, and extra productive. That’s the true untapped alternative with AI.

• Jon Whittle, Director, Data61, CSIRO

This text is republished from The Dialog below a Artistic Commons license. Learn the unique article.