As generative artificial intelligence (AI) advances at breakneck speed, it is upending assumptions about which jobs are \u201csafe\u201d from automation.<\/p>\n
Disruption now extends well beyond manual or routine work into white-collar roles once considered untouchable. Tools such as ChatGPT, Claude and Midjourney can produce policy briefs, analytical reports, software code, design assets and marketing copy in seconds.<\/p>\n
Even in specialised domains, systems such as PolicyPulse<\/a> can generate structured briefs and thematic syntheses \u2013 tasks that once required teams of experts.<\/p>\n If AI can so easily replicate large swaths of professional output, how much of the economy rests on work that creates the appearance of value rather than tangible impact?<\/p>\n And could New Zealand \u2013 anchored in sectors rooted in physical work, human judgement and essential services \u2013 be structurally better placed to thrive?<\/p>\n AI\u2019s exposure effect<\/p>\n A 2023 Goldman Sachs report<\/a> estimated generative AI could automate work equivalent to 300 million full-time jobs globally. The highest exposure is in administrative, legal and other information-heavy sectors.<\/p>\n In 2024, the International Monetary Fund warned<\/a> that economies reliant on high-skilled services \u2013 such as education, law and finance \u2013 face both job losses and rising inequality.<\/p>\n This echoes author David Graeber\u2019s concept of \u201cbullshit jobs<\/a>\u201d \u2013 roles that add little genuine value. Between 2000 and 2018, most net job growth<\/a> came from low productivity service sectors such as marketing, consulting and corporate administration. These are precisely the kinds of tasks AI can now perform in seconds.<\/p>\n Consultancy firm McKinsey estimates<\/a> 60\u201370% of activities in office support, customer service and professional services can be automated. The OECD has noted<\/a> routine information processing jobs face the greatest risk. AI is not only replacing roles \u2013 it is revealing how insubstantial many of them were.<\/p>\n Some argue finance illustrates this reality starkly: intended to allocate capital efficiently, the sector has expanded beyond its productive purpose. <\/p>\n Businessman Adair Turner famously called much of it \u201csocially useless<\/a>\u201d, while research<\/a> from the Bank for International Settlements found oversized financial sectors can stifle innovation by diverting talent from more productive areas.<\/p>\n Now, AI is automating functions<\/a> such as risk modelling, compliance and equity research, prompting a reassessment of the sector\u2019s true economic value.<\/p>\n New Zealand\u2019s real-economy advantage<\/p>\n New Zealand \u2013 often caricatured as a remote, agrarian outpost \u2013 may be structurally insulated from the worst of the AI shock. Roughly 70% of its exports<\/a> come from agriculture, horticulture, seafood and forestry.<\/p>\n Domestically, leading employment sectors<\/a> include aged care, physiotherapy, plumbing, childcare and early childhood education. <\/p>\n These roles require physical dexterity, sensory judgement and human empathy \u2013 skills AI cannot yet credibly replicate.<\/p>\n In an era when many advanced economies are over-invested in finance, bureaucracy and \u201cbullshit jobs\u201d, New Zealand\u2019s focus on tangible, value-producing work could be a strategic strength.<\/p>\n Innovation in these sectors is happening too<\/a>. Robotic milking systems have improved dairy efficiency and animal welfare, biosecurity monitoring safeguards exports, and forestry research is targeting carbon neutral timber.<\/p>\n If finance reveals how AI strips away illusions, higher education shows its disruptive power. Generative AI can now produce essays<\/a> credible enough to pass as human work.<\/p>\n The humanities tend to reward theoretical fluency and stylistic polish \u2013 areas where AI excels. By contrast, science, technology, engineering and mathematics \u2013 the so-called STEM subjects \u2013 demand precision, formal logic and testable hypotheses, which are harder for AI to mimic. OECD data<\/a> has shown STEM-related occupations face the lowest automation risk.<\/p>\n New Zealand\u2019s recent investment in STEM education<\/a> is timely. But it must be matched by support for primary and secondary teachers \u2013 roles grounded in mentorship and adaptive instruction<\/a>, which remain beyond AI\u2019s reach.<\/p>\n A global pivot<\/p>\n Service-heavy economies such as Singapore, Britain and parts of the United States face growing pressure to adapt<\/a>.<\/p>\n