{"course":{"slug":"agents-and-automation","title":"Agents and Automation","url":"https://addaly.com/learn/agents-and-automation","level":"intermediate","track":"build"},"lesson":{"slug":"what-work-is-automatable","title":"Which work is worth automating","url":"https://addaly.com/learn/agents-and-automation/what-work-is-automatable","position":4,"of":72,"minutes":8,"answer":"Automate high-volume, low-variance work whose errors are cheap and visible; the rest costs more in maintenance and supervision than it returns.","keyIdea":"Automate high-volume, low-variance work whose errors are cheap and visible; the rest costs more in maintenance and supervision than it returns."},"body":{"format":"markdown","text":"## Two axes decide most of it\n\nPut any task on two axes: how often it happens, and how much the inputs vary between instances.\n\n- **High volume, low variance.** Invoices from the same twelve suppliers. This is the sweet spot for plain code, and it is where automation pays.\n- **High volume, high variance.** Support emails from strangers. This is where a model earns its cost, and where you will need everything in the rest of this course.\n- **Low volume, low variance.** Runs four times a year. Write a checklist. The build cost will never come back.\n- **Low volume, high variance.** A person does this. Automating it is a hobby.\n\nMost failed automation projects are the bottom-right quadrant with a budget attached.\n\n## The arithmetic people skip\n\nA task takes fifteen minutes a week. That is thirteen hours a year. The build takes eight hours. Payback in about seven months, which sounds fine.\n\nNow add the term nobody writes down. Any automation touching an external API breaks around three or four times a year, and each break costs roughly an hour to diagnose and fix. Real saving: thirteen minus four, so nine hours a year against an eight-hour build plus your attention forever.\n\nThat is still positive, and it is much thinner than the pitch. Be honest about it, because the honest version still wins often, usually for a reason other than hours. Automation makes a task get done *consistently*, and *at all*, at 6am on the Monday after a long weekend. That is frequently worth more than the time.\n\n## Three questions about being wrong\n\nBefore you automate, price a mistake:\n\n1. **How visible is a wrong output?** A wrong number in a report someone reads is caught. A wrong tag on a record nobody opens is not.\n2. **How reversible is it?** A draft is free. A sent email is not. A payment is not.\n3. **How many happen before anyone notices?** This is the one that hurts. A 2% error rate is nothing at ten a week and a catastrophe at ten thousand a night.\n\nCheap, visible, reversible errors: automate freely and fix as you go. Expensive, invisible, irreversible: either do not, or put a person in the path, which is module four.\n\n## The tasks that look automatable and are not\n\n- **Work whose real content is an undocumented decision.** If nobody can write the rule down, you cannot encode it, and a model will guess at it convincingly.\n- **Work whose input format changes because a human keeps changing it.** You are not automating a process; you are chasing one.\n- **Work that exists so that somebody is accountable for having looked.** Automate the looking and you have removed the thing the work was for. The output was never the point; the reader was.\n\n## Instrument before you build\n\nAlmost everyone is wrong about how often the task happens. Log it for two weeks first: a line in a text file each time somebody does it, with how long it took. Two weeks of that has killed more bad automation projects than any amount of argument, and it gives you the denominator you will need later when you want to say whether the thing worked.\n\nThen automate the version you measured, not the version you remember."},"check":{"question":"A five-person team wants to automate a quarterly compliance sign-off. The task takes about two hours, four times a year. The rules change most quarters, and the sign-off exists so that a named manager has reviewed the figures. What is the best assessment?","options":[{"text":"Automate it with a fixed workflow, since the steps are the same each quarter.","correct":false,"tempting":"Treats stable steps as stable requirements; the rules change most quarters, so the workflow is wrong again by the time it next runs."},{"text":"Automate it with an agent, since the changing rules need judgement.","correct":false,"tempting":"Applies the most expensive machine to the least frequent task, and the build cost never returns across four runs a year."},{"text":"Do not automate it: eight hours a year of low-volume, high-variance work whose whole purpose is that a person looked.","correct":true},{"text":"Automate the data gathering but keep the sign-off manual, then extend it next year.","correct":false,"tempting":"The right instinct in general, but here the gathering is a small part of eight hours a year, so maintenance still exceeds the saving."}]},"prev":{"slug":"mapping-the-process-first","title":"Watch the process before you automate it","url":"https://addaly.com/learn/agents-and-automation/mapping-the-process-first"},"next":{"slug":"the-cost-of-being-wrong","title":"What a wrong answer costs, and who finds it","url":"https://addaly.com/learn/agents-and-automation/the-cost-of-being-wrong"},"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) 'Which work is worth automating', in Agents and Automation. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/agents-and-automation/what-work-is-automatable (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":"Which work is worth automating","courseTitle":"Agents and Automation","module":{"index":1,"slug":"deciding-what-to-build","title":"Before you build: script, workflow, or agent"},"answer":null,"keyIdea":"Automate high-volume, low-variance work whose errors are cheap and visible; the rest costs more in maintenance and supervision than it returns.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"two-axes-decide-most-of-it","runs":[{"text":"Two axes decide most of it"}]},{"t":"p","runs":[{"text":"Put any task on two axes: how often it happens, and how much the inputs vary between instances."}]},{"t":"ul","items":[[{"text":"High volume, low variance.","b":true},{"text":" Invoices from the same twelve suppliers. This is the sweet spot for plain code, and it is where automation pays."}],[{"text":"High volume, high variance.","b":true},{"text":" Support emails from strangers. This is where a model earns its cost, and where you will need everything in the rest of this course."}],[{"text":"Low volume, low variance.","b":true},{"text":" Runs four times a year. Write a checklist. The build cost will never come back."}],[{"text":"Low volume, high variance.","b":true},{"text":" A person does this. Automating it is a hobby."}]]},{"t":"p","runs":[{"text":"Most failed automation projects are the bottom-right quadrant with a budget attached."}]},{"t":"h2","id":"the-arithmetic-people-skip","runs":[{"text":"The arithmetic people skip"}]},{"t":"p","runs":[{"text":"A task takes fifteen minutes a week. That is thirteen hours a year. The build takes eight hours. Payback in about seven months, which sounds fine."}]},{"t":"p","runs":[{"text":"Now add the term nobody writes down. Any automation touching an external API breaks around three or four times a year, and each break costs roughly an hour to diagnose and fix. Real saving: thirteen minus four, so nine hours a year against an eight-hour build plus your attention forever."}]},{"t":"p","runs":[{"text":"That is still positive, and it is much thinner than the pitch. Be honest about it, because the honest version still wins often, usually for a reason other than hours. Automation makes a task get done "},{"text":"consistently","i":true},{"text":", and "},{"text":"at all","i":true},{"text":", at 6am on the Monday after a long weekend. That is frequently worth more than the time."}]},{"t":"h2","id":"three-questions-about-being-wrong","runs":[{"text":"Three questions about being wrong"}]},{"t":"p","runs":[{"text":"Before you automate, price a mistake:"}]},{"t":"ol","start":1,"items":[[{"text":"How visible is a wrong output?","b":true},{"text":" A wrong number in a report someone reads is caught. A wrong tag on a record nobody opens is not."}],[{"text":"How reversible is it?","b":true},{"text":" A draft is free. A sent email is not. A payment is not."}],[{"text":"How many happen before anyone notices?","b":true},{"text":" This is the one that hurts. A 2% error rate is nothing at ten a week and a catastrophe at ten thousand a night."}]]},{"t":"p","runs":[{"text":"Cheap, visible, reversible errors: automate freely and fix as you go. Expensive, invisible, irreversible: either do not, or put a person in the path, which is module four."}]},{"t":"h2","id":"the-tasks-that-look-automatable-and-are-not","runs":[{"text":"The tasks that look automatable and are not"}]},{"t":"ul","items":[[{"text":"Work whose real content is an undocumented decision.","b":true},{"text":" If nobody can write the rule down, you cannot encode it, and a model will guess at it convincingly."}],[{"text":"Work whose input format changes because a human keeps changing it.","b":true},{"text":" You are not automating a process; you are chasing one."}],[{"text":"Work that exists so that somebody is accountable for having looked.","b":true},{"text":" Automate the looking and you have removed the thing the work was for. The output was never the point; the reader was."}]]},{"t":"h2","id":"instrument-before-you-build","runs":[{"text":"Instrument before you build"}]},{"t":"p","runs":[{"text":"Almost everyone is wrong about how often the task happens. Log it for two weeks first: a line in a text file each time somebody does it, with how long it took. Two weeks of that has killed more bad automation projects than any amount of argument, and it gives you the denominator you will need later when you want to say whether the thing worked."}]},{"t":"p","runs":[{"text":"Then automate the version you measured, not the version you remember."}]}],"check":{"question":[{"text":"A five-person team wants to automate a quarterly compliance sign-off. The task takes about two hours, four times a year. The rules change most quarters, and the sign-off exists so that a named manager has reviewed the figures. What is the best assessment?"}],"options":[{"text":[{"text":"Automate it with a fixed workflow, since the steps are the same each quarter."}],"correct":false,"tempting":[{"text":"Treats stable steps as stable requirements; the rules change most quarters, so the workflow is wrong again by the time it next runs."}]},{"text":[{"text":"Automate it with an agent, since the changing rules need judgement."}],"correct":false,"tempting":[{"text":"Applies the most expensive machine to the least frequent task, and the build cost never returns across four runs a year."}]},{"text":[{"text":"Do not automate it: eight hours a year of low-volume, high-variance work whose whole purpose is that a person looked."}],"correct":true,"tempting":null},{"text":[{"text":"Automate the data gathering but keep the sign-off manual, then extend it next year."}],"correct":false,"tempting":[{"text":"The right instinct in general, but here the gathering is a small part of eight hours a year, so maintenance still exceeds the saving."}]}],"order":[1,3,2,0],"why":null},"prev":{"slug":"mapping-the-process-first","title":"Watch the process before you automate it","minutes":9},"next":{"slug":"the-cost-of-being-wrong","title":"What a wrong answer costs, and who finds it","minutes":9}}}