Approaches to regulating artificial intelligence (AI) have turned from discussions around safety, ethics and human rights, into ones around productivity, efficiency and worker outputs. This report is a literature review around current AI productivity claims. It looks at studies which provide insights and analysis from real-world use cases, are from independent, credible institutions, and consider a holistic understanding of the AI environment, including negative effects and harms. It is intended to provide key insights and evidence-based counterpoints to current claims around AI productivity.

The report recommends that further study and independent analysis is needed to qualify claims around the productivity benefits of AI, that negative externalities should not be ignored and instead be critical factors for consideration, and that AI is a national concern with diverse impacts requiring a holistic approach and not just filtered through a productivity lens.

Four main themes which emerge from the literature.

1. Productivity benefits are uneven, unclear and will likely take many years to materialise.
2. In many cases, AI is creating extra workload or generating new tasks as a result of its use.
3. Workers face new issues with AI use, including cognitive decline or debt, extra pressure and disadvantaging a new generation which could cause productivity issues in the medium to long term.
4. There are many unresolved issues – like worker displacement, unclear liabilities, privacy breaches and copyright breaches that remain unanswered and should be factored in.