{"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":"rules-beat-models","title":"When a plain rule beats a model","url":"https://addaly.com/learn/ai-for-small-business/rules-beat-models","position":56,"of":88,"minutes":9,"answer":"Anything that happens the same way every time is a rule, and rules are cheaper, instant, deterministic and impossible to talk into inventing something.","keyIdea":"Anything that happens the same way every time is a rule, and rules are cheaper, instant, deterministic and impossible to talk into inventing something, so build the rules first and add a model step only where the input varies and a person checks the output."},"body":{"format":"markdown","text":"## The question to ask of any repetitive task\n\nDoes this need judgement, or does it need doing?\n\nIf the same thing happens every time (an invoice raised when a job is marked complete, a reminder 24 hours before an appointment, a backup at midnight, a file copied when a form is submitted) that is a rule. Rules are cheap, instant, deterministic, auditable, and they cannot invent anything.\n\nPutting a model in a rule's place adds cost, latency, variation and a new failure mode, in exchange for nothing. This is the single most common way small businesses waste money on AI, and it is usually because the automation was sold as AI.\n\n## The tools, with honest pricing\n\n**Zapier**: the easiest, connects to nearly everything, free tier is small and paid tiers become expensive at volume.\n\n**Make**: similar, generally cheaper per operation, slightly steeper to learn.\n\n**n8n**: open source and free if you host it yourself, roughly USD 5 a month on a small server. More work to set up, no per-task charge, and your data stays where you put it.\n\n**Google Apps Script**: free, built into Google Workspace, runs on a schedule. If your business lives in Sheets and Gmail, this does an enormous amount for nothing.\n\n**Whatever your software already does.** Most accounting, booking and shop platforms have automation rules built in and switched off. Check before adding a vendor; this is the third time this course has said so, and it remains the most frequently skipped step.\n\n## A worked division\n\nA joinery workshop's enquiry-to-invoice flow:\n\n- Enquiry arrives → **rule**: create a record, notify the owner.\n- Classify the enquiry → **model**: this is meaning, and rules cannot do it.\n- Quote request → **rule**: create a quote from the template.\n- Draft the covering note → **model**: varying text.\n- Quote accepted → **rule**: create the job, order materials, diary the start date.\n- Job complete → **rule**: raise the invoice, send it, schedule reminders.\n- Payment late → **rule** to detect, **model** to draft the chaser at the right firmness.\n- Customer complains → **model** to draft, person to send.\n\nTwo model steps, six rule steps. That is a healthy ratio, and it is roughly what a working small-business automation looks like once the enthusiasm has settled.\n\n> Figure: A joinery workshop, marked rule or model at every step. Steps: 1. Enquiry arrives (Rule. Create the record, notify the owner. Instant and deterministic.); 2. Classify it (Model. This is meaning, and a keyword filter cannot do it.); 3. Quote requested (Rule builds it from the template; the model drafts the covering note; you set the price.); 4. Quote accepted (Rule. Create the job, order the materials, diary the start date.); 5. Job complete (Rule. Raise the invoice, send it, schedule the reminders.); 6. Payment late (Rule detects it. Model drafts the chaser at the right firmness. You press send.). Two model steps and six rule steps, which is roughly what a working small-business automation looks like once the enthusiasm has settled. Anything that happens the same way every time is a rule, and a rule cannot be talked into inventing something.\n\n## Where an AI step belongs in a workflow\n\nThree places, and they share a shape: the input varies, the output is checked, and nothing irreversible happens automatically.\n\n**Classification**: routing by meaning.\n**Extraction**: pulling fields out of a document into structured data.\n**Drafting**: producing text a person will read before it goes.\n\nIf an AI step in your workflow does something else (decides, sends, pays, publishes) look at it again. That is where module two's warning about hostile input becomes concrete, because an automated path that acts is a path somebody else's text can drive.\n\n## Build it so it tells you when it breaks\n\nAutomations fail silently, which is their characteristic failure. A vendor changes an interface, a token expires, a folder is renamed, and the reminders stop going out. Nobody notices for three weeks, because nothing produces an error where a person would see it.\n\nTwo cheap defences. Have the automation send you a weekly summary of what it did: a line saying \"47 reminders sent\" is enough to notice when it says nothing. And check anything money-related monthly against the source system.\n\n## The order of operations\n\nDo the rules first. They are cheaper, they work reliably, and they usually deliver more time back than the model steps do.\n\nThen add one model step, and only where the input varies. Businesses that do it the other way round end up with an expensive, occasionally creative system doing a job that a scheduled message did perfectly.\n\n## Start with one, and write down what it does\n\nThe failure pattern with automation tools is building six flows in an enthusiastic weekend, none documented, and discovering four months later that nobody knows what happens when a form is submitted.\n\nOne automation. Write down in plain words what triggers it, what it does, and what to check if it looks wrong. Keep that beside the prompt library from module two: same file, same discipline.\n\nThen live with it for a fortnight before building the second. The side effects of the first will teach you what the second should be, and they are never the ones you predicted.\n\n## Know how to turn it off\n\nBefore you switch anything on, know how to switch it off, and make sure somebody else knows too.\n\nThe bad afternoon in automation is the loop: a rule that creates a record that triggers a rule that creates a record. It is easy to build accidentally, it sends four hundred emails before anybody notices, and the only useful skill in that moment is knowing where the off switch is."},"check":{"question":"A shop's automation sends a review request after each purchase, using AI to write a personalised message each time. What is the strongest argument for replacing the AI step with a template?","options":[{"text":"Personalised messages risk containing details the customer did not share, which could appear intrusive and damage the relationship with buyers who value privacy.","correct":false,"tempting":"Imagines a privacy failure from data the automation would have to be given, when the more basic point is that nothing here varies enough to need generating."},{"text":"Templates can be tested once and then relied upon, whereas generated messages vary between customers and cannot be individually reviewed before sending at this volume.","correct":false,"tempting":"Gets close by naming the review problem, and stops short of the reason it matters: there is no variation in the task that justifies generating anything."},{"text":"The message is the same request every time, so it is a rule; generating it adds cost, latency and the possibility of saying something unintended for no benefit.","correct":true},{"text":"AI-written review requests may fall foul of platform rules on review solicitation, since some platforms restrict the wording that businesses may use when asking customers to leave feedback about a purchase.","correct":false,"tempting":"Invokes a platform restriction that applies to the wording whoever writes it, and would not be resolved by switching to a template."}]},"prev":{"slug":"forms-tax-and-officialdom","title":"Official letters, forms and the line at filing","url":"https://addaly.com/learn/ai-for-small-business/forms-tax-and-officialdom"},"next":{"slug":"working-with-your-accountant","title":"Arriving prepared, and what never to outsource","url":"https://addaly.com/learn/ai-for-small-business/working-with-your-accountant"},"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) 'When a plain rule beats a model', in AI for a Small Business. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-small-business/rules-beat-models (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":"When a plain rule beats a model","courseTitle":"AI for a Small Business","module":{"index":6,"slug":"money-and-the-back-office","title":"Money, paperwork and the back office"},"answer":null,"keyIdea":"Anything that happens the same way every time is a rule, and rules are cheaper, instant, deterministic and impossible to talk into inventing something, so build the rules first and add a model step only where the input varies and a person checks the output.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"flow","steps":[{"label":"Enquiry arrives","detail":"Rule. Create the record, notify the owner. Instant and deterministic."},{"label":"Classify it","detail":"Model. This is meaning, and a keyword filter cannot do it."},{"label":"Quote requested","detail":"Rule builds it from the template; the model drafts the covering note; you set the price."},{"label":"Quote accepted","detail":"Rule. Create the job, order the materials, diary the start date."},{"label":"Job complete","detail":"Rule. Raise the invoice, send it, schedule the reminders."},{"label":"Payment late","detail":"Rule detects it. Model drafts the chaser at the right firmness. You press send."}],"title":"A joinery workshop, marked rule or model at every step","caption":"Two model steps and six rule steps, which is roughly what a working small-business automation looks like once the enthusiasm has settled. Anything that happens the same way every time is a rule, and a rule cannot be talked into inventing something."},"alt":"Figure: A joinery workshop, marked rule or model at every step. Steps: 1. Enquiry arrives (Rule. Create the record, notify the owner. Instant and deterministic.); 2. Classify it (Model. This is meaning, and a keyword filter cannot do it.); 3. Quote requested (Rule builds it from the template; the model drafts the covering note; you set the price.); 4. Quote accepted (Rule. Create the job, order the materials, diary the start date.); 5. Job complete (Rule. Raise the invoice, send it, schedule the reminders.); 6. Payment late (Rule detects it. Model drafts the chaser at the right firmness. You press send.). Two model steps and six rule steps, which is roughly what a working small-business automation looks like once the enthusiasm has settled. Anything that happens the same way every time is a rule, and a rule cannot be talked into inventing something."}},"blocks":[{"t":"h2","id":"the-question-to-ask-of-any-repetitive-task","runs":[{"text":"The question to ask of any repetitive task"}]},{"t":"p","runs":[{"text":"Does this need judgement, or does it need doing?"}]},{"t":"p","runs":[{"text":"If the same thing happens every time (an invoice raised when a job is marked complete, a reminder 24 hours before an appointment, a backup at midnight, a file copied when a form is submitted) that is a rule. Rules are cheap, instant, deterministic, auditable, and they cannot invent anything."}]},{"t":"p","runs":[{"text":"Putting a model in a rule’s place adds cost, latency, variation and a new failure mode, in exchange for nothing. This is the single most common way small businesses waste money on AI, and it is usually because the automation was sold as AI."}]},{"t":"h2","id":"the-tools-with-honest-pricing","runs":[{"text":"The tools, with honest pricing"}]},{"t":"p","runs":[{"text":"Zapier","b":true},{"text":": the easiest, connects to nearly everything, free tier is small and paid tiers become expensive at volume."}]},{"t":"p","runs":[{"text":"Make","b":true},{"text":": similar, generally cheaper per operation, slightly steeper to learn."}]},{"t":"p","runs":[{"text":"n8n","b":true},{"text":": open source and free if you host it yourself, roughly USD 5 a month on a small server. More work to set up, no per-task charge, and your data stays where you put it."}]},{"t":"p","runs":[{"text":"Google Apps Script","b":true},{"text":": free, built into Google Workspace, runs on a schedule. If your business lives in Sheets and Gmail, this does an enormous amount for nothing."}]},{"t":"p","runs":[{"text":"Whatever your software already does.","b":true},{"text":" Most accounting, booking and shop platforms have automation rules built in and switched off. Check before adding a vendor; this is the third time this course has said so, and it remains the most frequently skipped step."}]},{"t":"h2","id":"a-worked-division","runs":[{"text":"A worked division"}]},{"t":"p","runs":[{"text":"A joinery workshop’s enquiry-to-invoice flow:"}]},{"t":"ul","items":[[{"text":"Enquiry arrives → "},{"text":"rule","b":true},{"text":": create a record, notify the owner."}],[{"text":"Classify the enquiry → "},{"text":"model","b":true},{"text":": this is meaning, and rules cannot do it."}],[{"text":"Quote request → "},{"text":"rule","b":true},{"text":": create a quote from the template."}],[{"text":"Draft the covering note → "},{"text":"model","b":true},{"text":": varying text."}],[{"text":"Quote accepted → "},{"text":"rule","b":true},{"text":": create the job, order materials, diary the start date."}],[{"text":"Job complete → "},{"text":"rule","b":true},{"text":": raise the invoice, send it, schedule reminders."}],[{"text":"Payment late → "},{"text":"rule","b":true},{"text":" to detect, "},{"text":"model","b":true},{"text":" to draft the chaser at the right firmness."}],[{"text":"Customer complains → "},{"text":"model","b":true},{"text":" to draft, person to send."}]]},{"t":"p","runs":[{"text":"Two model steps, six rule steps. That is a healthy ratio, and it is roughly what a working small-business automation looks like once the enthusiasm has settled."}]},{"t":"h2","id":"where-an-ai-step-belongs-in-a-workflow","runs":[{"text":"Where an AI step belongs in a workflow"}]},{"t":"p","runs":[{"text":"Three places, and they share a shape: the input varies, the output is checked, and nothing irreversible happens automatically."}]},{"t":"p","runs":[{"text":"Classification","b":true},{"text":": routing by meaning. "},{"text":"Extraction","b":true},{"text":": pulling fields out of a document into structured data. "},{"text":"Drafting","b":true},{"text":": producing text a person will read before it goes."}]},{"t":"p","runs":[{"text":"If an AI step in your workflow does something else (decides, sends, pays, publishes) look at it again. That is where module two’s warning about hostile input becomes concrete, because an automated path that acts is a path somebody else’s text can drive."}]},{"t":"h2","id":"build-it-so-it-tells-you-when-it-breaks","runs":[{"text":"Build it so it tells you when it breaks"}]},{"t":"p","runs":[{"text":"Automations fail silently, which is their characteristic failure. A vendor changes an interface, a token expires, a folder is renamed, and the reminders stop going out. Nobody notices for three weeks, because nothing produces an error where a person would see it."}]},{"t":"p","runs":[{"text":"Two cheap defences. Have the automation send you a weekly summary of what it did: a line saying “47 reminders sent” is enough to notice when it says nothing. And check anything money-related monthly against the source system."}]},{"t":"h2","id":"the-order-of-operations","runs":[{"text":"The order of operations"}]},{"t":"p","runs":[{"text":"Do the rules first. They are cheaper, they work reliably, and they usually deliver more time back than the model steps do."}]},{"t":"p","runs":[{"text":"Then add one model step, and only where the input varies. Businesses that do it the other way round end up with an expensive, occasionally creative system doing a job that a scheduled message did perfectly."}]},{"t":"h2","id":"start-with-one-and-write-down-what-it-does","runs":[{"text":"Start with one, and write down what it does"}]},{"t":"p","runs":[{"text":"The failure pattern with automation tools is building six flows in an enthusiastic weekend, none documented, and discovering four months later that nobody knows what happens when a form is submitted."}]},{"t":"p","runs":[{"text":"One automation. Write down in plain words what triggers it, what it does, and what to check if it looks wrong. Keep that beside the prompt library from module two: same file, same discipline."}]},{"t":"p","runs":[{"text":"Then live with it for a fortnight before building the second. The side effects of the first will teach you what the second should be, and they are never the ones you predicted."}]},{"t":"h2","id":"know-how-to-turn-it-off","runs":[{"text":"Know how to turn it off"}]},{"t":"p","runs":[{"text":"Before you switch anything on, know how to switch it off, and make sure somebody else knows too."}]},{"t":"p","runs":[{"text":"The bad afternoon in automation is the loop: a rule that creates a record that triggers a rule that creates a record. It is easy to build accidentally, it sends four hundred emails before anybody notices, and the only useful skill in that moment is knowing where the off switch is."}]}],"check":{"question":[{"text":"A shop’s automation sends a review request after each purchase, using AI to write a personalised message each time. What is the strongest argument for replacing the AI step with a template?"}],"options":[{"text":[{"text":"Personalised messages risk containing details the customer did not share, which could appear intrusive and damage the relationship with buyers who value privacy."}],"correct":false,"tempting":[{"text":"Imagines a privacy failure from data the automation would have to be given, when the more basic point is that nothing here varies enough to need generating."}]},{"text":[{"text":"Templates can be tested once and then relied upon, whereas generated messages vary between customers and cannot be individually reviewed before sending at this volume."}],"correct":false,"tempting":[{"text":"Gets close by naming the review problem, and stops short of the reason it matters: there is no variation in the task that justifies generating anything."}]},{"text":[{"text":"The message is the same request every time, so it is a rule; generating it adds cost, latency and the possibility of saying something unintended for no benefit."}],"correct":true,"tempting":null},{"text":[{"text":"AI-written review requests may fall foul of platform rules on review solicitation, since some platforms restrict the wording that businesses may use when asking customers to leave feedback about a purchase."}],"correct":false,"tempting":[{"text":"Invokes a platform restriction that applies to the wording whoever writes it, and would not be resolved by switching to a template."}]}],"order":[0,2,1,3],"why":null},"prev":{"slug":"forms-tax-and-officialdom","title":"Official letters, forms and the line at filing","minutes":9},"next":{"slug":"working-with-your-accountant","title":"Arriving prepared, and what never to outsource","minutes":8}}}