“Automation replaces tasks, not jobs. A job is a bundle of tasks, and when part of the bundle becomes cheap, value shifts to the part that was not automated — often rising rather than falling.”David Autor · Why Are There Still So Many Jobs? The History and Future of Workplace Automation · 2015 · David Autor, «Journal of Economic Perspectives», vol. 29, nr. 3, 2015 — redare a tezei sarcinilor, necolaționată pe articol
Tasks get automated, not jobs — the boundary runs inside every profession.
Autor showed across a century of data that each wave of automation abolished tasks and shifted value rather than emptying the labour market. The cash machine did not abolish the teller, it changed their bundle of tasks. What gets automated fast: anything with a verifiable answer, a repeated pattern and a well-defined input — routine translation, summaries, standard code, the first draft of any text, extracting data from a document, answering frequent questions. What stays expensive precisely because the rest gets cheap: one, defining the problem, that is knowing what is worth asking; two, judgement under insufficient data, where no verifiable answer exists; three, responsibility — someone has to sign, and a tool cannot be legally accountable; four, relationship and negotiation, where who says it matters and not only what is said; five, non-routine physical work in disorganised spaces, the hardest of all to automate; six, taste, that is knowing which of ten instantly produced variants is good. What this means for a child today: do not choose by what is fashionable, choose by how much of their future bundle falls in the second list. And an honest warning: nobody knows how fast the boundary moves, so the right bet is not on one occupation but on the ability to learn a second and a third quickly.
Autor pushes the idea further: the boundary does not split occupations, it splits people within the same occupation. An accountant who only enters invoices does something different from one who decides what an odd number means before signing the report. The first task bundle gets automated; the second becomes more valuable, because analysis is now cheap while the decision stays expensive. That is why two people with the same diploma can face opposite fates. The useful question for a child is not "which job", but "which part of the job will I be doing". Yet how do you find out which part stays human in a field you don't know yet?
Passage 2: How do you spot in practice which part of a job gets automated? Autor offers a simple test. Ask: does the task have a verifiable answer, a repeating pattern, and a well-defined input? If yes, it is a candidate for automation. A lawyer, for instance, does not do one thing: checking standard clauses is a repeating pattern, but deciding whether the client should sign at all is another story. The first part gets done by machine; the second stays with a person, because someone must carry responsibility for the signature. A tool cannot be taken to court. So when your child says they want a profession, unpack it with them into tasks and apply the test. But one question remains: what do you do when nobody knows where the line will fall in five years?
Why it matters Because parents today choose between 'computing' and 'law' as if these were stable categories, when the boundary runs through the inside of every occupation: within the same profession some people will do the automatable work and others the part that stays expensive.