Work and businesses that win on the way to AGI

Short version: the wellness picks hold up best. Premium entertainment is a spending story more than a jobs story. Coastal real estate is an asset story, not a career story.

Only weakly so far, in the official projections. The jobs that grow fastest are in the middle of the exposure range, not the bottom. Being hard for AI to do is not enough. You also need a reason for demand to grow.

How to read the exposure scores. BLS's new category combines five sources, from theory-based scores to observed AI use. BLS warns that exposure "does not imply job loss, productivity gains, automation probability, or wage effects." Eloundou et al.'s β is the share of a job's tasks an LLM could speed up by half or more. Felten et al.'s AIOE ranks occupations by the abilities that language models touch. All three are measured against today's AI, so in an AGI world they understate exposure for every job. What matters is the ranking: what AI reaches first.

Industries tell the business-owner version of the story. The fastest-growing ones serve ageing bodies, power the AI build-out, or sell in-person experiences. The shrinking ones sell words, paperwork or labour that can be offshored.

Business types that fit the pattern. These are ranked by evidence strength, not by advice. (1) Ageing and care: home health, assisted living, outpatient and therapy clinics. They have the biggest projected job gains and the clearest demand driver, demographics. (2) Hands-on wellness: massage, physio, aesthetics, fitness studios, wellness travel. Fast growth, low exposure and low startup cost, but fragmented and often low-wage. (3) Experience venues and promoters: live events, sport, hospitality at the premium end. Spending is strong, but pricing power sits with a few large players. (4) Picks and shovels for the build-out: electricians, power-line work, HVAC, data-centre services. (5) Scarce places: prime real estate, an asset play covered in AGI assets.

The fourth pick is really about owning scarce, desirable places rather than working in real estate. Knight Frank's Prime International Residential Index (PIRI 100) is the standard benchmark for luxury homes across 100 markets.

What would make coastal prime a better AGI bet than a city flat? Fixed supply of beachfront, wealth concentrating among owners of capital (see AGI assets), and remote work letting the wealthy live where they like. Against it: climate risk and rising insurance costs on coasts, taxes aimed at the wealthy (Los Angeles' Measure ULA cooled super-prime sales there), and the fact that 1-year returns lag cities as often as they lead.

These candidates are not on the original list. They do at least as well in the projections and share the same logic: physical, local or trust-heavy work with a growing source of demand.

The losers share a profile: the work is mostly reading, writing, keying data or following a script at a desk or phone. BLS rates almost all of these High or Very high AI exposure, and many were shrinking before ChatGPT.

Surprises worth noting. Some exposed 'word' jobs are not projected to shrink much yet: writers −0.3%, paralegals −0.3%, interpreters +2.0%. BLS projects trends, and exposure has so far shown up in early-career hiring more than in headcount (Anthropic's −14% job-finding rate for 22-25-year-olds). If AGI arrives inside the decade, these are the rows to watch. Their shrinkage would also feed the debt and unrest risks in AI and the great divergence.