Economy
Intelligence becomes abundant. That sounds great until you ask who owns it, who gets the value, and what happens to everyone else.
The core problem nobody has solved
Value is created. The question is who captures it.
The economic history of the last 300 years is a story of productivity gains eventually distributing themselves through wages. The industrial revolution destroyed jobs, created different jobs, and over decades, living standards rose. This pattern gave us a comfortable narrative: technology destroys, then creates, and everyone ends up better off.
AGI breaks this pattern in a fundamental way. Previous technologies replaced physical labour or augmented cognitive labour. AGI replaces cognitive labour itself. When intelligence — the ability to plan, decide, create, strategise — becomes abundant and effectively free, the mechanism by which most people earn their income collapses.
You can't retrain for "jobs that don't exist yet" when the thing that creates new jobs — intelligence — is now available at near-zero marginal cost. The buffer that protected workers during previous technological shifts was human cognition. That buffer is disappearing.
This is why the economic implications of AGI are so much larger than most people realise. It's not the next industrial revolution. It's something fundamentally different.
The numbers
These numbers from 2023-24 research are already conservative. The capability jumps since then suggest higher figures across all three metrics. But the numbers matter less than the shape: the gains are concentrated, the losses are distributed, and the policy framework to handle this mismatch barely exists anywhere in the world.
Two scenarios
The range of possible outcomes is enormous. Policy choices determine which one we get.
The optimistic case: superabundance
AI generates unprecedented wealth. The cost of goods and services collapses — food, housing, energy, healthcare all become dramatically cheaper as AI optimises every supply chain. Work becomes optional for those who don't want it, and more rewarding for those who do. Universal Basic Income covers essentials. Most people earn additional income through AI-augmented work, entrepreneurship, or creative pursuits. The economic pie grows so fast that distributional conflicts are manageable. Tax revenue from AI productivity funds a robust social safety net. The transition is messy but the destination is genuinely better.
The pessimistic case: concentration
AI gains flow predominantly to capital owners — the companies that build the models, the investors who funded them, the shareholders who own them. The middle class hollows out as AI replaces mid-skill cognitive work. Low-end service jobs survive longer but wages stagnate. UBI, if implemented, covers survival but not mobility or aspiration. Wealth concentration reaches levels last seen in the Gilded Age. Social unrest grows. Democracy strains under the pressure of mass economic displacement. The "retrain" narrative becomes a cruel joke as people discover AI can do the retrained job too.
Reality will fall somewhere between these. Which side we lean toward depends entirely on policy decisions made in the next 3-5 years.
Universal Basic Income: theory to reality
UBI has been debated for decades. It may become not just desirable but necessary.
How it could work
Every adult receives a regular cash payment, unconditional and sufficient to cover basic needs. Estimated cost for NZ at $1K/month per adult: roughly $30-40 billion/year — a large but not impossible share of GDP. Funded through a combination of automation taxes, data dividends, carbon taxes, and reduced welfare administration costs.
Global pilots
Finland's experiment (2017-18) showed recipients reported better wellbeing and slightly higher employment. Kenya's ongoing trial (GiveDirectly) shows economic activity increases, not decreases. US cities (Stockton, California) found recipients were more likely to find full-time work. Canada's "Mincome" experiment (1970s) showed reduced hospitalisation rates and no significant work reduction.
The evidence so far contradicts the "UBI makes people lazy" narrative. People use the money to invest in education, start businesses, relocate for better opportunities, and care for family members. The concern isn't laziness — it's cost and political will.
Who pays
New Zealand context
NZ is a small, open economy heavily exposed to global technology shifts. Here's what it means specifically.
New Zealand's economy is 65% services — finance, tourism, retail, healthcare, education. Most of these sectors have high cognitive labour components that AI will transform. The country's geographical isolation means it can't compete on manufacturing scale. Its competitive advantages — agriculture, tourism, a clean brand, a stable legal system, abundant renewable energy — are real but don't insulate it from the labour market shifts AI brings.
Estimated impact: 300,000+ NZ jobs significantly exposed to AI automation by 2035. Clerical, administrative, and customer service roles are most vulnerable. But NZ also has structural advantages — a flexible labour market, a relatively small population that could adapt faster, and the ability to pilot new social policies.
Current policy: No UBI. NZ Super is universal for 65+. ACC covers accidents but not general income loss. The government has established an AI Strategy Taskforce under MBIE. The conversation about AI and work is still at the "awareness building" stage, not the "policy design" stage. The window for proactive policy is closing as the technology accelerates.
Key question: Will NZ tax AI productivity gains to fund income redistribution, or let market forces determine the outcome? The choice shapes the country's social structure for decades.
Jobs: most and least exposed
Most exposed
- Administrative, clerical, data entry — AI handles with >80% accuracy
- Customer service, call centres — tier-1 support is already AI-handled
- Retail and checkout — cashierless stores are proven technology
- Basic coding and content writing — AI produces production-quality output
- Translation and transcription — near-human quality for common languages
- Accounting clerks and bookkeeping — software eats the routine work
- Legal document review and discovery — AI reads faster, finds more
- Insurance underwriting and claims processing — pattern recognition at scale
Least exposed
- Healthcare bedside (nursing, physiotherapy) — physical presence required
- Skilled trades (electrician, plumber, mechanic) — unstructured environments
- Creative direction and strategy — human judgment, vision, taste
- Leadership and high-stakes decision-making — accountability can't be delegated
- Therapy, coaching, social work — empathy and human connection
- Early childhood education — care and relationship-building
- Emergency services (fire, police, paramedic) — physical crisis response
- Some construction and agriculture — varied, messy, physical environments
"Least exposed" doesn't mean immune. It means these roles are further from the automation frontier. Each year that frontier moves. The trades look safer than admin today. In 10 years, that gap may narrow significantly.
What to do
For individuals
Build transferable skills, especially human interaction ones. Develop AI fluency — not just using tools, but understanding what they can and can't do. Diversify your income sources if possible. Invest in relationships and networks — they're the most durable career asset. Stay adaptable and avoid over-investing in narrow specialisation.
For policy
Support UBI pilots at national and local levels. Tax automation revenue — capital gains, corporate profits, compute usage. Invest in transition programmes that genuinely prepare people for AI-augmented work, not 20th century retraining. Build sovereign wealth funds from AI productivity taxes. Create the policy infrastructure now, before the transition becomes a crisis.
The bottom line
The economic transformation AGI brings is larger than the industrial revolution, larger than the internet, larger than globalisation. It changes the fundamental relationship between human labour and economic value. That's not hyperbole — it's a structural analysis of what happens when cognitive labour becomes abundant.
The optimistic case is real: massive wealth creation, cheaper goods, more human freedom. The pessimistic case is also real: extreme concentration, social breakdown, a world divided between AI owners and everyone else. The difference is policy. And policy is made by people who understand what's coming.
Value is created. The question is who captures it. The answer is being decided right now.