{"course":{"slug":"ai-careers","title":"Working in AI","url":"https://addaly.com/learn/ai-careers","level":"beginner","track":"work"},"lesson":{"slug":"the-long-game","title":"The second job, the plateau, and the ten-year view","url":"https://addaly.com/learn/ai-careers/the-long-game","position":83,"of":85,"minutes":8,"answer":"A plateau is a signal to change the work, not to work harder, and the durable assets are a domain, thirty people who have seen you work, and the habit of finishing.","keyIdea":"A plateau is a signal to change the work, not to work harder, and the durable assets are a domain, thirty people who have seen you work, and the habit of finishing."},"body":{"format":"markdown","text":"## The pattern most careers follow\n\nYear one you learn the tools and feel permanently behind. Year two you become useful and the panic subsides. Year three you either deepen, broaden, or quietly stall.\n\nThe stall is the dangerous one because it feels like competence. Work is easy, you are respected, nothing is difficult. From the inside it is indistinguishable from having arrived.\n\n## Diagnosing a plateau, concretely\n\nThree questions, asked honestly once a year:\n\n- **Could I have done last quarter's work eighteen months ago?** If yes, you have been paid to repeat yourself.\n- **Is there anybody here I am learning from?** If the answer is no, no amount of effort fixes it, because skill at this level is mostly absorbed from people.\n- **Has anything I built been used by more people this year than last?** Scope growing is the clearest external signal that you are.\n\nThree noes is not a crisis. It is a year's notice.\n\n## The three moves, and what each costs\n\n**Deeper.** A specialism: data platform engineering, ML systems, a regulated domain. Costs: narrower job market, higher pay within it, and a real risk if the specialism fades.\n\n**Wider.** Product, management, solutions work. Costs: your technical depth erodes within about two years and going back is hard. Worth doing when you want it, not as an escape from a technical problem.\n\n**Out.** Contracting, a small consultancy, your own product. Costs: income variance, the selling half of the job, and no colleagues. Freedom that people romanticise and then find lonely.\n\nDoing nothing is a fourth option, chosen by default rather than decided. Money and comfort compound, the cost of the plateau is invisible, and it presents its bill at the next interview, when you discover what your market rate actually is.\n\n## Gaps, caring years and non-linear paths\n\nIf you took two years out (for a child, a parent, illness, a country change) the advice is one plain sentence with no apology, then talk about the work. *\"I was out of work for two years caring for my father. I kept my SQL current on a project for a local charity.\"* No elaboration, no defensiveness.\n\nEmployers who dwell disproportionately on a gap are telling you how they will treat any future human circumstance of yours. That is worth learning during the interview rather than after.\n\n## Burnout, in this field specifically\n\nData and AI work has three ingredients that produce it reliably. Pages at three in the morning, because pipelines run at night. A permanent, manufactured sense of being behind, from the previous lesson. And no natural end state: the data is never clean, the model is never finished, and there is always one more thing to check.\n\nTwo guards that actually work, both structural rather than attitudinal:\n\n- **An on-call arrangement that is written down**, with a rota and a definition of what warrants waking someone. An informal expectation that you are always reachable is the most common cause of people leaving this work altogether.\n- **A definition of done for your own projects.** Written before you start. Without one, every project expands until you stop from exhaustion rather than completion.\n\n## The asset almost nobody builds deliberately\n\nThree months of expenses saved changes which jobs you can refuse, which requests you can decline, and how a negotiation feels. It is the most useful career asset available to most people and it is never described as one, because it is boring and it is not a skill.\n\nIf you are early and underpaid, this is hard, and it is still worth naming as the goal rather than pretending the choice is only about learning.\n\n## What survives ten years\n\nNot the tools. Every framework named in this course will be replaced, and the replacements will be replaced.\n\nWhat survives:\n\n- **A domain.** Ten years of understanding how hospitals, freight, lending or agriculture actually work is not reproducible by anyone in a hurry.\n- **About thirty people who have seen you work** and would take your call. This is what a network is, stripped of its unpleasant connotations, and it is built by being useful to people one step behind you, for years.\n- **A written record of what you built and decided.** The habit from the case study lesson, kept up for a decade, is a career's evidence that no employer owns.\n- **The habit of finishing.** Rarer than talent and more valuable.\n\n## The limitation\n\nNone of this is a guarantee. Markets contract, sectors are hit, well-run companies fail, and a good decision can still end badly through nothing you did. Career advice that promises otherwise is describing survivors.\n\nWhat the evidence habit actually buys is not immunity. It is that your work is legible to somebody else when you need it to be, which is the difference between a bad year and a lost one."},"check":{"question":"Three years into a data job, Sofia is respected, comfortable, and cannot remember the last time work was difficult. Her manager is pleased with her. What is the most accurate reading?","options":[{"text":"She has reached competence in the role, and the sensible move is to stay and consolidate.","correct":false,"tempting":"Reads ease as mastery, which is exactly how the plateau disguises itself, and consolidation of unchanging work is repetition."},{"text":"She is likely underpaid relative to the market and should test it by interviewing elsewhere.","correct":false,"tempting":"Possibly true and beside the point; the issue is what she is learning, and a higher salary for identical work extends the plateau rather than ending it."},{"text":"Ease is the signal to check three things (whether she could have done this work eighteen months ago, whether anyone there is teaching her anything, and whether her scope has grown) and to change the work rather than work harder if the answers are no.","correct":true},{"text":"Her manager's satisfaction confirms she is performing well, so the discomfort she feels is misplaced.","correct":false,"tempting":"Uses her manager's view as the measure, when a manager is generally pleased by exactly the reliability that a plateau produces."}]},"prev":{"slug":"the-refusals","title":"The things you will be asked to build","url":"https://addaly.com/learn/ai-careers/the-refusals"},"next":{"slug":"writing-for-people-who-decide","title":"The one-pager that gets a decision","url":"https://addaly.com/learn/ai-careers/writing-for-people-who-decide"},"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) 'The second job, the plateau, and the ten-year view', in Working in AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-careers/the-long-game (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":"The second job, the plateau, and the ten-year view","courseTitle":"Working in AI","module":{"index":9,"slug":"staying","title":"Staying: the first ninety days and the ten years after"},"answer":null,"keyIdea":"A plateau is a signal to change the work, not to work harder, and the durable assets are a domain, thirty people who have seen you work, and the habit of finishing.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"the-pattern-most-careers-follow","runs":[{"text":"The pattern most careers follow"}]},{"t":"p","runs":[{"text":"Year one you learn the tools and feel permanently behind. Year two you become useful and the panic subsides. Year three you either deepen, broaden, or quietly stall."}]},{"t":"p","runs":[{"text":"The stall is the dangerous one because it feels like competence. Work is easy, you are respected, nothing is difficult. From the inside it is indistinguishable from having arrived."}]},{"t":"h2","id":"diagnosing-a-plateau-concretely","runs":[{"text":"Diagnosing a plateau, concretely"}]},{"t":"p","runs":[{"text":"Three questions, asked honestly once a year:"}]},{"t":"ul","items":[[{"text":"Could I have done last quarter’s work eighteen months ago?","b":true},{"text":" If yes, you have been paid to repeat yourself."}],[{"text":"Is there anybody here I am learning from?","b":true},{"text":" If the answer is no, no amount of effort fixes it, because skill at this level is mostly absorbed from people."}],[{"text":"Has anything I built been used by more people this year than last?","b":true},{"text":" Scope growing is the clearest external signal that you are."}]]},{"t":"p","runs":[{"text":"Three noes is not a crisis. It is a year’s notice."}]},{"t":"h2","id":"the-three-moves-and-what-each-costs","runs":[{"text":"The three moves, and what each costs"}]},{"t":"p","runs":[{"text":"Deeper.","b":true},{"text":" A specialism: data platform engineering, ML systems, a regulated domain. Costs: narrower job market, higher pay within it, and a real risk if the specialism fades."}]},{"t":"p","runs":[{"text":"Wider.","b":true},{"text":" Product, management, solutions work. Costs: your technical depth erodes within about two years and going back is hard. Worth doing when you want it, not as an escape from a technical problem."}]},{"t":"p","runs":[{"text":"Out.","b":true},{"text":" Contracting, a small consultancy, your own product. Costs: income variance, the selling half of the job, and no colleagues. Freedom that people romanticise and then find lonely."}]},{"t":"p","runs":[{"text":"Doing nothing is a fourth option, chosen by default rather than decided. Money and comfort compound, the cost of the plateau is invisible, and it presents its bill at the next interview, when you discover what your market rate actually is."}]},{"t":"h2","id":"gaps-caring-years-and-non-linear-paths","runs":[{"text":"Gaps, caring years and non-linear paths"}]},{"t":"p","runs":[{"text":"If you took two years out (for a child, a parent, illness, a country change) the advice is one plain sentence with no apology, then talk about the work. "},{"text":"”I was out of work for two years caring for my father. I kept my SQL current on a project for a local charity.”","i":true},{"text":" No elaboration, no defensiveness."}]},{"t":"p","runs":[{"text":"Employers who dwell disproportionately on a gap are telling you how they will treat any future human circumstance of yours. That is worth learning during the interview rather than after."}]},{"t":"h2","id":"burnout-in-this-field-specifically","runs":[{"text":"Burnout, in this field specifically"}]},{"t":"p","runs":[{"text":"Data and AI work has three ingredients that produce it reliably. Pages at three in the morning, because pipelines run at night. A permanent, manufactured sense of being behind, from the previous lesson. And no natural end state: the data is never clean, the model is never finished, and there is always one more thing to check."}]},{"t":"p","runs":[{"text":"Two guards that actually work, both structural rather than attitudinal:"}]},{"t":"ul","items":[[{"text":"An on-call arrangement that is written down","b":true},{"text":", with a rota and a definition of what warrants waking someone. An informal expectation that you are always reachable is the most common cause of people leaving this work altogether."}],[{"text":"A definition of done for your own projects.","b":true},{"text":" Written before you start. Without one, every project expands until you stop from exhaustion rather than completion."}]]},{"t":"h2","id":"the-asset-almost-nobody-builds-deliberately","runs":[{"text":"The asset almost nobody builds deliberately"}]},{"t":"p","runs":[{"text":"Three months of expenses saved changes which jobs you can refuse, which requests you can decline, and how a negotiation feels. It is the most useful career asset available to most people and it is never described as one, because it is boring and it is not a skill."}]},{"t":"p","runs":[{"text":"If you are early and underpaid, this is hard, and it is still worth naming as the goal rather than pretending the choice is only about learning."}]},{"t":"h2","id":"what-survives-ten-years","runs":[{"text":"What survives ten years"}]},{"t":"p","runs":[{"text":"Not the tools. Every framework named in this course will be replaced, and the replacements will be replaced."}]},{"t":"p","runs":[{"text":"What survives:"}]},{"t":"ul","items":[[{"text":"A domain.","b":true},{"text":" Ten years of understanding how hospitals, freight, lending or agriculture actually work is not reproducible by anyone in a hurry."}],[{"text":"About thirty people who have seen you work","b":true},{"text":" and would take your call. This is what a network is, stripped of its unpleasant connotations, and it is built by being useful to people one step behind you, for years."}],[{"text":"A written record of what you built and decided.","b":true},{"text":" The habit from the case study lesson, kept up for a decade, is a career’s evidence that no employer owns."}],[{"text":"The habit of finishing.","b":true},{"text":" Rarer than talent and more valuable."}]]},{"t":"h2","id":"the-limitation","runs":[{"text":"The limitation"}],"role":"limit"},{"t":"p","runs":[{"text":"None of this is a guarantee. Markets contract, sectors are hit, well-run companies fail, and a good decision can still end badly through nothing you did. Career advice that promises otherwise is describing survivors."}]},{"t":"p","runs":[{"text":"What the evidence habit actually buys is not immunity. It is that your work is legible to somebody else when you need it to be, which is the difference between a bad year and a lost one."}]}],"check":{"question":[{"text":"Three years into a data job, Sofia is respected, comfortable, and cannot remember the last time work was difficult. Her manager is pleased with her. What is the most accurate reading?"}],"options":[{"text":[{"text":"She has reached competence in the role, and the sensible move is to stay and consolidate."}],"correct":false,"tempting":[{"text":"Reads ease as mastery, which is exactly how the plateau disguises itself, and consolidation of unchanging work is repetition."}]},{"text":[{"text":"She is likely underpaid relative to the market and should test it by interviewing elsewhere."}],"correct":false,"tempting":[{"text":"Possibly true and beside the point; the issue is what she is learning, and a higher salary for identical work extends the plateau rather than ending it."}]},{"text":[{"text":"Ease is the signal to check three things (whether she could have done this work eighteen months ago, whether anyone there is teaching her anything, and whether her scope has grown) and to change the work rather than work harder if the answers are no."}],"correct":true,"tempting":null},{"text":[{"text":"Her manager’s satisfaction confirms she is performing well, so the discomfort she feels is misplaced."}],"correct":false,"tempting":[{"text":"Uses her manager’s view as the measure, when a manager is generally pleased by exactly the reliability that a plateau produces."}]}],"order":[3,2,0,1],"why":null},"prev":{"slug":"the-refusals","title":"The things you will be asked to build","minutes":8},"next":{"slug":"writing-for-people-who-decide","title":"The one-pager that gets a decision","minutes":9}}}