Research findings try to sway debate over AI's role in boosting labour productivity

There’s been increasing political focus on productivity, especially in Australia, where growth is at a historic low. The government is exploring the role of digital technologies like AI in improving output. While the tech sector promotes AI’s potential, recent research casts doubt on its real productivity impact.

A piece in The Conversation unpacks this, starting with definitions: productivity measures how efficiently inputs like labour or materials generate outputs. It exists on three levels—individual, organisational, and national. Although national productivity aggregates the others, it’s difficult to trace how personal or business-level gains affect broader GDP per hour worked.

Emerging studies on AI’s effect at the individual level are inconclusive. One 2025 study at Procter & Gamble showed solo AI users matched the output of two-person teams without AI. A 2023 BCG study found AI sped up consultant tasks by 18%. In another 2023 study, a Fortune 500 firm’s customer agents saw a 14% average improvement, but gains were much higher—35%—for novices. Yet AI’s impact isn’t universally positive.

A survey of 2,500 professionals found AI increased workloads for 77% of respondents; nearly half weren’t sure how to use it for productivity. Barriers included the need to fact-check AI, upskill workers, and overcome unrealistic expectations. In a CSIRO study of 300 government employees using Microsoft 365 Copilot, most reported productivity gains—but 30% didn’t, and even satisfied users said the gains were smaller than expected.

At the organisational level, AI’s impact remains hard to isolate, as many social and structural variables also shape outcomes. The OECD estimates productivity gains from traditional, task-specific AI at just 0–11%. A 2022 review of 300,000 U.S. firms showed no major link between AI use and productivity, though robotics and cloud computing did correlate with higher output. Explanations range from slow adoption to challenges in separating AI’s role from other factors.

Sometimes AI use even creates extra labour—training, monitoring, or fixing AI systems—which obscures benefits. National productivity, meanwhile, remains largely untouched by AI. As with earlier tech, AI might need time before its effects fully register, with infrastructure and skills acting as bottlenecks.

But history shows results aren’t guaranteed. While the internet boosted productivity, the jury’s still out on mobile phones and social media, whose gains vary by industry. The popular view that AI boosts efficiency by automating repetitive tasks and freeing creative time oversimplifies how work happens. As the article points out, faster isn’t always better—clearing email quicker can just mean more messages flood in. True productivity sometimes requires slowing down to think, reflect, and eventually innovate.

By Nazrin Sadigova