{"course":{"slug":"ai-actually-explained","title":"AI, Actually Explained","url":"https://addaly.com/learn/ai-actually-explained","level":"beginner","track":"start"},"lesson":{"slug":"six-trades-honestly","title":"Six trades, as they stood in 2025","url":"https://addaly.com/learn/ai-actually-explained/six-trades-honestly","position":81,"of":92,"minutes":10,"answer":"In software, writing, support, translation, illustration and law, the same three things happened by 2025 (the routine task moved to the model, the entry rung thinned.","keyIdea":"In software, writing, support, translation, illustration and law, the same three things happened by 2025 (the routine task moved to the model, the entry rung thinned, and at least one firm that replaced people reversed course) with the details differing by how far demand stretched."},"body":{"format":"markdown","text":"## One pattern, six times\n\nThis lesson is a survey, not a forecast. For six trades that this course's readers are most likely to be in or heading for, it records what had actually happened by 2025, one or two concrete facts each, and where each sat on the elasticity rule.\n\n## Software\n\nMore code was being written than ever and fewer people were being hired to write it. Postings for software jobs on the largest job boards fell by a third or more from their 2022 peak; graduate hiring at large technology firms fell further. The tools were everywhere (a large share of new code at the biggest firms was model-drafted by 2025), and the productivity evidence was, as the earlier lessons showed, split by task and by seniority. What changed in the job was the ratio of writing to reviewing. Demand for software is the stretching kind, and the honest position is that the fall in hiring has causes beyond the tools and that nobody has yet seen whether cheap code means far more software or just fewer people.\n\n## Writing and marketing\n\nThe freelance data showed the fall first. In-house, content teams shrank and output rose: the same person now produced several times the copy by editing drafts. A sports magazine was caught in late 2023 publishing articles under invented author names with generated headshots, and a wave of generated filler across the web followed. What held value was reporting, voice, and a name that a reader trusted. The product description is gone as a human task. The column is not.\n\n## Customer support\n\nThe most-cited case reversed itself. In early 2024 a payments company announced that its assistant was handling the work of seven hundred agents and that it had stopped hiring. In 2025 its chief executive said the quality had suffered and that it was hiring people again, for the cases the assistant handled badly. The experiment from earlier in this module had measured the real effect: fewer, better-supported agents, with the tool handling the routine and the person the rest. Demand for support does not stretch, so the occupation shrinks by the routine share; the reversal shows the routine share was smaller than the announcement assumed.\n\n## Translation\n\nThe trade with the longest exposure. Machine translation had been taking the low-value end since 2016, and the surveys of 2024 recorded translators losing work and income in large numbers. The remaining human work moved toward checking machine drafts, at lower rates. What held was the high-stakes end (legal, medical, literary, anything where an error costs more than the translation) and the languages where the models remain weak, which includes most Indian ones. Elasticity is the open question: if cheap translation means everything gets translated, the occupation grows around the checking; so far it has not.\n\n## Illustration and design\n\nThe freelance fall was steepest here, around a sixth of postings. Working illustrators reported lost commissions in the 2024 surveys; a tabletop-games publisher faced a public backlash in 2024 over generated art in its products and changed its policy. What held was art direction, brand, character and anything a client needed to own cleanly: the copyright uncertainty from a later module became, for illustrators, a selling point. The free tools from the beyond-text module are the same tools the trade now uses.\n\n## Law and accounting\n\nDocument review, standard drafting and research were the junior's tasks and are the model's. Large firms rolled the tools out across 2024 and 2025; the exposure studies put law near the top of the list. The fake-citations case from 2023 (six invented precedents, a fine) became the trade's cautionary tale, and the practitioner's signature became the product. Junior intake at some firms fell; at others the training changed. The rule that a licensed person must be accountable holds the ladder up here more than in any other trade, and it is the clearest example of a profession whose structure resists the entry-rung problem.\n\n## What the six have in common\n\nThe routine task moved to the model. The entry rung thinned. At least one prominent firm in each trade overclaimed, and at least one reversed. The value moved toward accountability, judgement, and what a client needs to own. And in every case the people who used the tools well did better than the people who did not, which is the last lesson's subject.\n\n## What they do not have in common\n\nHow far demand stretches. Software and perhaps translation can grow around the tool. Support and stock illustration cannot. Writing and law sit between. When you place your own trade on this list, that is the question to ask first."},"check":{"question":"A company announced in 2024 that its assistant did the work of 700 support agents; in 2025 it said it was hiring people again. Which reading fits the evidence in this module?","options":[{"text":"The assistant stopped working as the model was updated.","correct":false,"tempting":"Blames a model change; the reversal was about quality on the cases the assistant had always handled badly."},{"text":"The routine share of support work was smaller than the announcement assumed, and the rest needed people.","correct":true},{"text":"Customers refused to talk to a machine, so the company gave up on the tool.","correct":false,"tempting":"Overstates the reversal; the tool stayed in place for the routine cases and people were added for the rest."},{"text":"The company found that demand for support grew once it was cheaper to provide, so it needed more agents to meet it.","correct":false,"tempting":"Applies the stretching case to the trade where it does not apply; nobody wants more support tickets."}]},"prev":{"slug":"the-hands-that-trained-it","title":"The labour that trained it","url":"https://addaly.com/learn/ai-actually-explained/the-hands-that-trained-it"},"next":{"slug":"what-to-do-about-it","title":"What an early-career person can actually do","url":"https://addaly.com/learn/ai-actually-explained/what-to-do-about-it"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Six trades, as they stood in 2025', in AI, Actually Explained. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-actually-explained/six-trades-honestly (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Six trades, as they stood in 2025","courseTitle":"AI, Actually Explained","module":{"index":8,"slug":"ai-and-jobs","title":"AI and work, with the evidence"},"answer":null,"keyIdea":"In software, writing, support, translation, illustration and law, the same three things happened by 2025 (the routine task moved to the model, the entry rung thinned, and at least one firm that replaced people reversed course) with the details differing by how far demand stretched.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"one-pattern-six-times","runs":[{"text":"One pattern, six times"}]},{"t":"p","runs":[{"text":"This lesson is a survey, not a forecast. For six trades that this course’s readers are most likely to be in or heading for, it records what had actually happened by 2025, one or two concrete facts each, and where each sat on the elasticity rule."}]},{"t":"h2","id":"software","runs":[{"text":"Software"}]},{"t":"p","runs":[{"text":"More code was being written than ever and fewer people were being hired to write it. Postings for software jobs on the largest job boards fell by a third or more from their 2022 peak; graduate hiring at large technology firms fell further. The tools were everywhere (a large share of new code at the biggest firms was model-drafted by 2025), and the productivity evidence was, as the earlier lessons showed, split by task and by seniority. What changed in the job was the ratio of writing to reviewing. Demand for software is the stretching kind, and the honest position is that the fall in hiring has causes beyond the tools and that nobody has yet seen whether cheap code means far more software or just fewer people."}]},{"t":"h2","id":"writing-and-marketing","runs":[{"text":"Writing and marketing"}]},{"t":"p","runs":[{"text":"The freelance data showed the fall first. In-house, content teams shrank and output rose: the same person now produced several times the copy by editing drafts. A sports magazine was caught in late 2023 publishing articles under invented author names with generated headshots, and a wave of generated filler across the web followed. What held value was reporting, voice, and a name that a reader trusted. The product description is gone as a human task. The column is not."}]},{"t":"h2","id":"customer-support","runs":[{"text":"Customer support"}]},{"t":"p","runs":[{"text":"The most-cited case reversed itself. In early 2024 a payments company announced that its assistant was handling the work of seven hundred agents and that it had stopped hiring. In 2025 its chief executive said the quality had suffered and that it was hiring people again, for the cases the assistant handled badly. The experiment from earlier in this module had measured the real effect: fewer, better-supported agents, with the tool handling the routine and the person the rest. Demand for support does not stretch, so the occupation shrinks by the routine share; the reversal shows the routine share was smaller than the announcement assumed."}]},{"t":"h2","id":"translation","runs":[{"text":"Translation"}]},{"t":"p","runs":[{"text":"The trade with the longest exposure. Machine translation had been taking the low-value end since 2016, and the surveys of 2024 recorded translators losing work and income in large numbers. The remaining human work moved toward checking machine drafts, at lower rates. What held was the high-stakes end (legal, medical, literary, anything where an error costs more than the translation) and the languages where the models remain weak, which includes most Indian ones. Elasticity is the open question: if cheap translation means everything gets translated, the occupation grows around the checking; so far it has not."}]},{"t":"h2","id":"illustration-and-design","runs":[{"text":"Illustration and design"}]},{"t":"p","runs":[{"text":"The freelance fall was steepest here, around a sixth of postings. Working illustrators reported lost commissions in the 2024 surveys; a tabletop-games publisher faced a public backlash in 2024 over generated art in its products and changed its policy. What held was art direction, brand, character and anything a client needed to own cleanly: the copyright uncertainty from a later module became, for illustrators, a selling point. The free tools from the beyond-text module are the same tools the trade now uses."}]},{"t":"h2","id":"law-and-accounting","runs":[{"text":"Law and accounting"}]},{"t":"p","runs":[{"text":"Document review, standard drafting and research were the junior’s tasks and are the model’s. Large firms rolled the tools out across 2024 and 2025; the exposure studies put law near the top of the list. The fake-citations case from 2023 (six invented precedents, a fine) became the trade’s cautionary tale, and the practitioner’s signature became the product. Junior intake at some firms fell; at others the training changed. The rule that a licensed person must be accountable holds the ladder up here more than in any other trade, and it is the clearest example of a profession whose structure resists the entry-rung problem."}]},{"t":"h2","id":"what-the-six-have-in-common","runs":[{"text":"What the six have in common"}]},{"t":"p","runs":[{"text":"The routine task moved to the model. The entry rung thinned. At least one prominent firm in each trade overclaimed, and at least one reversed. The value moved toward accountability, judgement, and what a client needs to own. And in every case the people who used the tools well did better than the people who did not, which is the last lesson’s subject."}]},{"t":"h2","id":"what-they-do-not-have-in-common","runs":[{"text":"What they do not have in common"}]},{"t":"p","runs":[{"text":"How far demand stretches. Software and perhaps translation can grow around the tool. Support and stock illustration cannot. Writing and law sit between. When you place your own trade on this list, that is the question to ask first."}]}],"check":{"question":[{"text":"A company announced in 2024 that its assistant did the work of 700 support agents; in 2025 it said it was hiring people again. Which reading fits the evidence in this module?"}],"options":[{"text":[{"text":"The assistant stopped working as the model was updated."}],"correct":false,"tempting":[{"text":"Blames a model change; the reversal was about quality on the cases the assistant had always handled badly."}]},{"text":[{"text":"The routine share of support work was smaller than the announcement assumed, and the rest needed people."}],"correct":true,"tempting":null},{"text":[{"text":"Customers refused to talk to a machine, so the company gave up on the tool."}],"correct":false,"tempting":[{"text":"Overstates the reversal; the tool stayed in place for the routine cases and people were added for the rest."}]},{"text":[{"text":"The company found that demand for support grew once it was cheaper to provide, so it needed more agents to meet it."}],"correct":false,"tempting":[{"text":"Applies the stretching case to the trade where it does not apply; nobody wants more support tickets."}]}],"order":[0,2,1,3],"why":null},"prev":{"slug":"the-hands-that-trained-it","title":"The labour that trained it","minutes":9},"next":{"slug":"what-to-do-about-it","title":"What an early-career person can actually do","minutes":9}}}