A merchandiser spends much of the working day producing and checking text: emails to a buyer, notes on a specification, a summary of an order's status, a reminder about an approaching deadline. An AI tool that can draft, summarise or reorganise text is now part of that daily admin work for many people in the trade, including on export orders like Order A. Used without understanding what the tool is actually doing, it becomes one more source of an unchecked error added to the order record. Used with a clear, working idea of what it is and is not, it becomes a genuine time saving on the routine writing tasks that surround every order.
This lesson gives you a plain, working explanation of what an AI tool does when it produces an answer, why that explains both its real strengths and its real limits, and why the merchandiser, not the tool, stays responsible for what goes out under their name. Without this understanding, a person either avoids a genuinely useful tool out of caution, or trusts an output that was never checked against the order's own facts.
You need: no previous lesson; a basic understanding of a merchandiser's daily tasks on an export order is useful but not required.
By the end of this lesson you can:
Not covered here: how any specific AI product works internally, or which AI tool to buy or subscribe to. Lesson 1.2 covers the tasks AI tends to do well; lesson 1.3 covers the mistakes it still makes.
| Term | Meaning |
|---|---|
| AI tool | Software that produces text, summaries or answers by predicting likely, well-formed output from patterns in the data it was built on. |
| Prompt | The instruction or question a person gives an AI tool to get a useful response. |
| AI-generated draft | Text an AI tool has produced that has not yet been checked against the facts it should match. |
| Spot-check | Comparing one specific detail in an output against a fact you already know is correct. |
| Human accountable decision | The principle that a person, not the AI tool, remains responsible for a work decision or a document sent under their name. |
| Data privacy | Keeping confidential business information out of a tool or a request where it should not be shared. |
An AI tool works by predicting what a well-formed answer to a given instruction is likely to look like, based on patterns in a very large amount of text it was built on. This is why it can draft a clear email, tidy a messy paragraph or produce a plausible-sounding summary in seconds. It is also why the tool has no direct knowledge of Order A's own agreed FOB, Delta Garments' own production schedule, or any other fact specific to your business, unless that fact is given to it in the request or checked afterwards against a source you trust. Treat every AI-generated draft the way you would treat a first draft from a new team member: useful as a starting point, not yet ready to send.
| What can differ | The shapes you will meet | Why it matters to you |
|---|---|---|
| Export order under Make to Order, factory side | A factory merchandiser like Mona uses an AI tool mainly to draft or tidy written communication about a confirmed order, such as a status update to a buyer. | The order's own facts (quantity, FOB, dates) are fixed and known, so the person's own check is quick: does the draft match what was actually agreed. |
| Domestic replenishment under Make to Stock | A domestic brand's planning team may ask an AI tool to summarise a stack of routine reorder notes into one short update. | Routine, repeated tasks are exactly where a fluent-but-unverified summary is most tempting to send unchecked, because it looks routine too. |
| Hybrid order handled through an agent | An agent working across several buyers' hybrid order mixes may use an AI tool to draft a comparison of two buyers' requirements. | Mixing facts from more than one buyer in one request raises the chance the tool blends details that must stay separate, unless the person checks each fact against its own source. |
| A new user compared with an experienced user | A person new to a role may not yet know which facts on an order are fixed and must be checked, while an experienced merchandiser already knows Order A's own figures by heart. | Recognising which facts are fixed and checkable is what makes the check quick; without that knowledge, a new user has no fast way to spot a wrong detail. |
| Different product families | Socks, knit-to-garment, terry and home-textile orders each carry their own fixed technical facts (gauge, GSM, pile height, set components) that an AI tool cannot know unless told. | The underlying rule is the same across every product family: supply the facts, then check the output against them. |
Delta Garments is producing Order A for Northline Apparel: 15,000 T-shirts, single jersey 180 GSM, Ne 30/1, agreed FOB USD 3.20 per piece, total order value USD 48,000. Mona, the merchandiser, needs to send Sara a short weekly update confirming that cutting has started on schedule and that no issue has come up.
Mona types the order's own facts into an AI tool: the order number, the quantity, the stage reached, and the date. The tool returns a clear, well-organised paragraph that reads exactly like a professional status email. Mona reads it against what she actually knows: cutting did start on the agreed date, the quantity mentioned matches 15,000 pieces, and no fact has been invented that she did not supply. Only after that check does she send it to Sara.
The worked numberThe real saving is 5 minutes, not the 10 minutes it might look like if the checking step is skipped. Skipping the check does not save the other 5 minutes; it only hides the risk of sending a wrong fact to Sara at Northline Apparel, which would cost far more than 5 minutes to correct once noticed.
Fluent, confident-sounding text is not the same as checked text. A common mistake is sending an AI-generated update, cost note or buyer email straight from the tool because it sounds professional, without comparing its specific facts, such as a quantity or a date, against what is actually true on the order. The cost of that mistake is not the tool's fault; it is the missing few minutes of the person's own check, and it lands on the person whose name is on the email, not on the tool.
Think of an AI tool as a fast, well-spoken assistant who has read an enormous amount of writing but was not in the room when your order was agreed. Ask it to write, tidy or summarise something, and it will do that quickly and clearly. Ask it to tell you a fact about your own order without giving it that fact first, and it may still answer confidently, even if the answer is wrong. The tool is good at sounding right. You are still the one who checks that it is right.
Mona is busy and asks an AI tool to "write a short message to the buyer confirming the ship date for Order A," without pasting in the actual current schedule from her own planning sheet.
Should Mona send the AI tool's message as it stands? Decide before opening the answer.
No. Mona must check the stated ship date against the production plan before sending anything to Sara.
1. Mona never gave the AI tool Order A's real ship date, so any date in the draft was predicted from general patterns, not read from the factory's own plan. 2. A wrong ship date sent to a buyer is a commercial fact, not a style choice, and Sara at Northline Apparel would reasonably treat it as confirmed once it arrives in writing. 3. The fix costs Mona only the time to open her own planning sheet and correct one sentence, far less than the time needed to correct a buyer's expectation after a wrong date has already been sent.
The AI tool did exactly what it is built to do: produce a fluent, plausible message. It was never the tool's task to know Order A's real schedule; that was always Mona's own fact to supply and check.
Order A's total order value is fixed at:
A wrong ship date, quantity or price sent to Northline Apparel because an AI-generated draft was never checked does not usually change this total figure directly, but it can trigger a real cost elsewhere: a rebooked shipment, a buyer's lost confidence in the factory's own updates, or hours spent correcting a written record once the wrong fact has already been read and acted on. Against an order worth USD 48,000, the cost of skipping a five-minute check is disproportionate to the time it was meant to save.
Use the T18-C01 AI use worksheet. In the Task log sheet, record each task you gave an AI tool, whether it needed you to supply the order's own facts, and what you checked in the output before using it. Start with Mona's own status-update example, then repeat the exercise for a task from your own week.
[Download the AI use worksheet](../../templates/T18-C01-ai-use-worksheet.xlsx)
1. Why does Mona's five-minute check of the AI-generated status update matter, even though the draft reads as well-written and professional?
2. A planning assistant at Cedar & Co. asks an AI tool to summarise this week's reorder notes without supplying the actual current stock figures. The summary states a stock number that sounds reasonable. What should the assistant do before acting on it?
| EN | AR | FR | TR | Meaning |
|---|---|---|---|---|
| AI tool | أداة ذكاء اصطناعي | Outil d'IA | Yapay zekâ aracı | Software that produces text, summaries or answers by predicting likely, well-formed output from patterns in the data it was built on. |
| Prompt | التوجيه | Consigne | Yönerge | The instruction or question a person gives an AI tool to get a useful response. |
| AI-generated draft | مسودة أنشأها الذكاء الاصطناعي | Brouillon généré par l'IA | Yapay zekânın oluşturduğu taslak | Text an AI tool has produced that has not yet been checked against the facts it should match. |
| Spot-check | تدقيق سريع | Vérification ponctuelle | Nokta kontrolü | Comparing one specific detail in an output against a fact you already know is correct. |
| Human accountable decision | قرار يتحمل مسؤوليته إنسان | Décision sous responsabilité humaine | İnsanın sorumlu olduğu karar | The principle that a person, not the AI tool, remains responsible for a work decision or a document sent under their name. |
| Data privacy | خصوصية البيانات | Confidentialité des données | Veri gizliliği | Keeping confidential business information out of a tool or a request where it should not be shared. |
Next you will learn which parts of a buyer's own daily work an AI tool tends to help with, and why those tasks suit it, seen through Sara's own work at Northline Apparel on Order A. Back to the course map