Robotics vs. Generative AI: Why They Exhibit Differing Labour Market Effects in Australia

Abstract

This paper analyses the differing labour market effects created by industrial robotics and generative Artificial Intelligence (AI) within the Australian economy. While industrial robotics and generative AI are often aggregated under the singular banner term of “automation”, treating them as a homogenous phenomenon downplays their radically divergent transmission mechanisms. By utilising a task-based framework, this paper argues that these two technologies propagate via entirely distinct tracks within the Australian economy; automation is not one overall phenomenon, but rather has a multifaceted dual-track impact, stemming from the automation of individual tasks as opposed to whole jobs. Track 1 investigates the substitution forces of robotics within sectors like mining and manufacturing. The capital intensive and highly structured nature of robotics implementation creates concentrated labour market displacement shocks that are felt within local communities. Track 2 examines the geographically diffused augmentation effects of generative AI across the services sector, specifically finance and retail. Driven by low-friction, decentralised implementation, generative AI has introduced an uneven wave of task transformation and restructuring as opposed to systemic job elimination. Via a qualitative review of domestic sectoral data, this study demonstrates that productivity gains realised at a national macro-level fail to capture the localised socioeconomic disruptions and adjustments observed at the community and firm level. By disaggregating these technological mechanisms, this paper leverages Australia’s distinct sectoral landscape to clarify the globally evolving friction between technological replacement and human reinstatement. In relation to controlling externalities, this paper uses the same disaggregation to conclude that because automation is not a singular process, labour market adaptation cannot be a singular response either.

Read full paper (PDF) →