{"course":{"slug":"ai-for-small-business","title":"AI for a Small Business","url":"https://addaly.com/learn/ai-for-small-business","level":"beginner","track":"work"},"lesson":{"slug":"automating-away-the-noticing","title":"The thing you lose that nobody counts","url":"https://addaly.com/learn/ai-for-small-business/automating-away-the-noticing","position":67,"of":88,"minutes":8,"answer":"Routine work carries incidental noticing (the odd invoice, the customer who has gone quiet) and automating the work removes the noticing unless you deliberately rebuild it somewhere else.","keyIdea":"Routine work carries incidental noticing (the odd invoice, the customer who has gone quiet) and automating the work removes the noticing unless you deliberately rebuild it somewhere else."},"body":{"format":"markdown","text":"## What the boring job was also doing\n\nSomeone in the business types every invoice, or reads every enquiry, or does the bank reconciliation on a Friday. The task is dull and it is the first candidate for automation.\n\nIt is also the reason you know that a supplier's prices crept up 4% in March, that a particular customer has not ordered since June, that two enquiries this month mentioned a product you stopped making, and that one invoice looked wrong in a way nobody can articulate.\n\nNone of that is in the job description. It is a by-product of a person touching every item, and it disappears the week the touching stops.\n\n## Why it is not on the spreadsheet\n\nThe two-week trial from module one measures hours saved and errors introduced. It cannot measure the thing you would have noticed and did not, because there is no record of a thought nobody had.\n\nSo the loss shows up late and disguised: a supplier overcharging for four months, a customer lost without a conversation, a product problem that reaches a review before it reaches you. Each one gets attributed to something else. The trial still reads as a success, and it may well be one: the point is that the ledger is incomplete, not that automation is a mistake.\n\n## Rebuilding the noticing on purpose\n\nThree cheap habits, and the first is the one that matters.\n\n**Look at the aggregate.** You have lost the per-item view; replace it with a summary that a person reads. Once a week: how many, of what kinds, biggest and smallest, anything that failed validation. This is the digest the log lesson makes possible, and reading it takes four minutes.\n\n**Keep a manual pass.** Once a month, do twenty by hand. Not a sample for quality control (though it serves that too) but to stay in contact with what the work actually looks like now. People who do this reliably find something in the first hour: a category that has grown, a phrasing customers have started using, a step that is no longer necessary.\n\n**Watch the edges, not the middle.** Automation handles the typical case beautifully and the edges badly, and the edges are where the information is. A queue of everything the automation declined to handle is the most interesting document in the business.\n\n## The sentence that should make you look again\n\n\"It has been running fine for months, so we stopped checking it.\"\n\nThat is two claims, and only one of them is supported. The automation has not produced a complaint for months, which is what \"fine\" actually means. Whether it has been producing quiet, tolerable, slightly wrong output the whole time is a question nobody has asked, because the person who would have noticed no longer sees the items.\n\nThe cost of finding out is an hour. The cost of not finding out accrues silently and surfaces as something else: a drop in repeat orders, a supplier relationship that went cold, a category of enquiry that stopped converting.\n\n## The deskilling question\n\nThere is a real effect and it is worth naming without drama. If the only person who could write your quotes now approves quotes written by a model, that skill decays, in them and in whoever they would have trained.\n\nStudies of automation in other trades (aviation, medicine, navigation) consistently find that supervising a system is a different and weaker skill than doing the job, and that the supervisor's ability to take over degrades with time. There is no reason to think office work is exempt.\n\nFor a small business the practical form of this is: what happens the week the tool is down, the vendor is acquired, or the price triples. If the answer is that nobody in the building can now do the work at the old speed, you have a continuity problem and not merely a subscription.\n\nThe mitigation is cheap and unpopular: the monthly manual pass again, and written procedures from module six that describe the work rather than the tool.\n\n## Where this leaves you\n\nAutomate the typical case. Read the aggregate. Touch the work by hand often enough to keep knowing what it is.\n\nThe businesses that get long-term value from this are not the ones that automated most. They are the ones that stayed in contact with their own operations while automating, which costs about an hour a month and is the first thing dropped when the month is busy."},"check":{"question":"A bakery automates supplier-invoice entry at rung four. Six months later it discovers a supplier's unit price rose twice without notice. Which mitigation would most likely have caught it, and why?","options":[{"text":"A weekly summary showing totals and unit prices by supplier, because the change is visible in the aggregate even though no single invoice looks wrong.","correct":true},{"text":"A validation rule rejecting any invoice whose total exceeds the previous one from the same supplier, so the automation halts and asks a person the first time a price moves.","correct":false,"tempting":"Would catch a rise but fires on every larger order too, so it is abandoned within a fortnight for noise."},{"text":"A monthly manual pass over twenty invoices chosen at random, which keeps somebody in contact with what the paperwork actually contains.","correct":false,"tempting":"A genuinely good habit, and a sample of twenty across all suppliers is unlikely to put two invoices from the same one side by side."},{"text":"Moving invoice entry back to rung three so that a person approves each one, restoring the per-item attention that entering them by hand used to provide.","correct":false,"tempting":"Restores the touching, and an approver checking that fields match the document still has no reason to compare against last quarter."}]},"prev":{"slug":"the-ai-employee-pitch","title":"The AI employee pitch, and the five questions that deflate it","url":"https://addaly.com/learn/ai-for-small-business/the-ai-employee-pitch"},"next":{"slug":"the-data-you-already-have","title":"What you are already sitting on, and what of it is usable","url":"https://addaly.com/learn/ai-for-small-business/the-data-you-already-have"},"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 thing you lose that nobody counts', in AI for a Small Business. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-small-business/automating-away-the-noticing (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 thing you lose that nobody counts","courseTitle":"AI for a Small Business","module":{"index":7,"slug":"letting-it-act","title":"Letting it act: automation you can trust with the keys"},"answer":null,"keyIdea":"Routine work carries incidental noticing (the odd invoice, the customer who has gone quiet) and automating the work removes the noticing unless you deliberately rebuild it somewhere else.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"what-the-boring-job-was-also-doing","runs":[{"text":"What the boring job was also doing"}]},{"t":"p","runs":[{"text":"Someone in the business types every invoice, or reads every enquiry, or does the bank reconciliation on a Friday. The task is dull and it is the first candidate for automation."}]},{"t":"p","runs":[{"text":"It is also the reason you know that a supplier’s prices crept up 4% in March, that a particular customer has not ordered since June, that two enquiries this month mentioned a product you stopped making, and that one invoice looked wrong in a way nobody can articulate."}]},{"t":"p","runs":[{"text":"None of that is in the job description. It is a by-product of a person touching every item, and it disappears the week the touching stops."}]},{"t":"h2","id":"why-it-is-not-on-the-spreadsheet","runs":[{"text":"Why it is not on the spreadsheet"}]},{"t":"p","runs":[{"text":"The two-week trial from module one measures hours saved and errors introduced. It cannot measure the thing you would have noticed and did not, because there is no record of a thought nobody had."}]},{"t":"p","runs":[{"text":"So the loss shows up late and disguised: a supplier overcharging for four months, a customer lost without a conversation, a product problem that reaches a review before it reaches you. Each one gets attributed to something else. The trial still reads as a success, and it may well be one: the point is that the ledger is incomplete, not that automation is a mistake."}]},{"t":"h2","id":"rebuilding-the-noticing-on-purpose","runs":[{"text":"Rebuilding the noticing on purpose"}]},{"t":"p","runs":[{"text":"Three cheap habits, and the first is the one that matters."}]},{"t":"p","runs":[{"text":"Look at the aggregate.","b":true},{"text":" You have lost the per-item view; replace it with a summary that a person reads. Once a week: how many, of what kinds, biggest and smallest, anything that failed validation. This is the digest the log lesson makes possible, and reading it takes four minutes."}]},{"t":"p","runs":[{"text":"Keep a manual pass.","b":true},{"text":" Once a month, do twenty by hand. Not a sample for quality control (though it serves that too) but to stay in contact with what the work actually looks like now. People who do this reliably find something in the first hour: a category that has grown, a phrasing customers have started using, a step that is no longer necessary."}]},{"t":"p","runs":[{"text":"Watch the edges, not the middle.","b":true},{"text":" Automation handles the typical case beautifully and the edges badly, and the edges are where the information is. A queue of everything the automation declined to handle is the most interesting document in the business."}]},{"t":"h2","id":"the-sentence-that-should-make-you-look-again","runs":[{"text":"The sentence that should make you look again"}]},{"t":"p","runs":[{"text":"“It has been running fine for months, so we stopped checking it.”"}]},{"t":"p","runs":[{"text":"That is two claims, and only one of them is supported. The automation has not produced a complaint for months, which is what “fine” actually means. Whether it has been producing quiet, tolerable, slightly wrong output the whole time is a question nobody has asked, because the person who would have noticed no longer sees the items."}]},{"t":"p","runs":[{"text":"The cost of finding out is an hour. The cost of not finding out accrues silently and surfaces as something else: a drop in repeat orders, a supplier relationship that went cold, a category of enquiry that stopped converting."}]},{"t":"h2","id":"the-deskilling-question","runs":[{"text":"The deskilling question"}]},{"t":"p","runs":[{"text":"There is a real effect and it is worth naming without drama. If the only person who could write your quotes now approves quotes written by a model, that skill decays, in them and in whoever they would have trained."}]},{"t":"p","runs":[{"text":"Studies of automation in other trades (aviation, medicine, navigation) consistently find that supervising a system is a different and weaker skill than doing the job, and that the supervisor’s ability to take over degrades with time. There is no reason to think office work is exempt."}]},{"t":"p","runs":[{"text":"For a small business the practical form of this is: what happens the week the tool is down, the vendor is acquired, or the price triples. If the answer is that nobody in the building can now do the work at the old speed, you have a continuity problem and not merely a subscription."}]},{"t":"p","runs":[{"text":"The mitigation is cheap and unpopular: the monthly manual pass again, and written procedures from module six that describe the work rather than the tool."}]},{"t":"h2","id":"where-this-leaves-you","runs":[{"text":"Where this leaves you"}]},{"t":"p","runs":[{"text":"Automate the typical case. Read the aggregate. Touch the work by hand often enough to keep knowing what it is."}]},{"t":"p","runs":[{"text":"The businesses that get long-term value from this are not the ones that automated most. They are the ones that stayed in contact with their own operations while automating, which costs about an hour a month and is the first thing dropped when the month is busy."}]}],"check":{"question":[{"text":"A bakery automates supplier-invoice entry at rung four. Six months later it discovers a supplier’s unit price rose twice without notice. Which mitigation would most likely have caught it, and why?"}],"options":[{"text":[{"text":"A weekly summary showing totals and unit prices by supplier, because the change is visible in the aggregate even though no single invoice looks wrong."}],"correct":true,"tempting":null},{"text":[{"text":"A validation rule rejecting any invoice whose total exceeds the previous one from the same supplier, so the automation halts and asks a person the first time a price moves."}],"correct":false,"tempting":[{"text":"Would catch a rise but fires on every larger order too, so it is abandoned within a fortnight for noise."}]},{"text":[{"text":"A monthly manual pass over twenty invoices chosen at random, which keeps somebody in contact with what the paperwork actually contains."}],"correct":false,"tempting":[{"text":"A genuinely good habit, and a sample of twenty across all suppliers is unlikely to put two invoices from the same one side by side."}]},{"text":[{"text":"Moving invoice entry back to rung three so that a person approves each one, restoring the per-item attention that entering them by hand used to provide."}],"correct":false,"tempting":[{"text":"Restores the touching, and an approver checking that fields match the document still has no reason to compare against last quarter."}]}],"order":[1,0,2,3],"why":null},"prev":{"slug":"the-ai-employee-pitch","title":"The AI employee pitch, and the five questions that deflate it","minutes":9},"next":{"slug":"the-data-you-already-have","title":"What you are already sitting on, and what of it is usable","minutes":9}}}