The standard account of AI and work assumes a formal job, a contract, an employer investing in tools, and a role made mostly of documents. That describes a real group of people and a small share of the world's workers.
Change any one of those assumptions and the conclusions change with it, often to the opposite.
What varies, and why it changes the answer
Four things move the outcome more than the technology does.
- How much work is informal. Where most work has no contract, the questions about employer-provided tools and retraining programmes simply do not apply, and the effects arrive through prices and demand instead.
- Where the work sits in a value chain. Economies built on delivering knowledge work to clients elsewhere are exposed very differently from economies where that work is consumed domestically.
- Demographics. An economy with a large young population entering work has an entry problem when junior tasks are automated. An economy with an ageing workforce has a transfer problem, which is nearly the opposite.
- Infrastructure and cost. Access to reliable connectivity and the price of these tools relative to local wages determines whether adoption is a decision or not an option.
Two examples that break the standard narrative
The advice most commonly given, move up the value chain into judgement work, assumes the judgement roles exist locally. Where an economy's knowledge work is largely delivered to clients abroad, the judgement roles are frequently held abroad too, and moving up the chain means changing employer or country rather than changing skills.
The reassurance that physical and interpersonal work is safe reads very differently where such work is informal, unprotected and poorly paid. Being unautomatable is not the same as being secure, and treating the two as equivalent is a comfortable mistake made from a distance.
What travels and what does not
Some of this generalises. Some of it is local and gets presented as universal, which is how people end up following advice written for a situation they are not in.
- Travels: judgement becoming more valuable than production, output no longer proving capability, and the need to demonstrate capability directly.
- Does not travel: assumptions about employer-funded retraining, about formal contracts, about which roles exist locally, and about whether adoption is optional.
- Worth checking before acting on: whether the roles the advice points toward actually exist in your market, and who holds them.
Why this matters for anybody writing about this
It is easy to write about the future of work as though there is one. There is not, and the version that gets written is usually the one belonging to the people doing the writing.
The honest position is to be specific about which situation a claim applies to, and to say so. That is less quotable and considerably more useful to somebody reading it from a different one.