An AI tool cannot be held accountable for a business decision, because it cannot be asked to correct its own mistake, cannot be paid, and cannot answer to a buyer or a factory owner. When AI-assisted work goes wrong, someone has to answer for it, and if that person was never named in advance, the business spends time arguing about whose mistake it was instead of fixing it. This is true whether the AI tool drafted a reply, calculated a figure, or helped write a material specification.
This lesson mainly teaches the decision flow for assigning accountability in advance: one named person owns each AI-assisted decision, and that person's accountability does not change because an AI tool was involved in producing it. Without this rule fixed ahead of time, "the AI got it wrong" becomes an answer no one can act on, while the real cost of the mistake still has to be paid by someone.
You need: Lesson 5.2's picture of the risks an ungoverned AI tool creates and Lesson 5.3's method for checking a tool before it is adopted. This lesson covers what happens once a tool is already in use and something goes wrong.
By the end of this lesson you can:
Not covered here: the policy that sets which tools may be used (Lesson 5.1) and the checks run before a tool is adopted (Lesson 5.3); this lesson assumes a tool is already approved and in use.
| Term | Meaning |
|---|---|
| Accountable person | The one named individual who answers for a business decision, whether or not an AI tool helped produce it. |
| Attribution error | Blaming an AI tool for a mistake instead of naming the accountable person who used its output without checking it. |
| Sign-off | The accountable person's confirmation that an AI-assisted output has been checked and may be acted on. |
| Correction record | A written note of what went wrong, who found it, and what was done to fix it, kept once an AI-assisted error is confirmed. |
| Escalation | Passing a decision or a found error up to a more senior or more accountable person when it is beyond the current person's authority to resolve alone. |
An AI tool can draft, calculate or recommend, but it cannot be the accountable party for a business decision, because accountability requires the ability to explain a decision, correct it, and answer to whoever is affected by it. The person who used the AI tool's output, or the person who approved that use, remains exactly as accountable as if they had produced the same output entirely by hand. Naming this person in advance, for each kind of AI-assisted decision, is what prevents "the AI got it wrong" from becoming the end of the conversation instead of the start of the correction.
This does not mean every small AI-assisted task needs a formal sign-off; it means every decision that could cost real money, time or trust has one named person whose job it is to check the AI tool's output before it is acted on, and to answer for it afterwards.
| What can differ | The shapes you will meet | Why it matters to you |
|---|---|---|
| Export order under Make to Order | An AI-drafted reply confirming a price or date to the buyer is signed and sent by a merchandiser. | The merchandiser who sends it is accountable, not the junior team member who drafted it with the AI tool, unless the merchandiser is also the one who typed the request. |
| Domestic replenishment under Make to Stock | An AI-recommended reorder quantity is approved and placed by a planning manager. | The planning manager who places the order is accountable for checking the recommendation, even if a more junior planner ran the AI tool. |
| Hybrid order handled through a mill | An AI-assisted material specification is used across a hybrid mix of order types for more than one buyer. | The mill's own technical manager who approves the specification for release is accountable for each buyer's own order, even if the same AI-assisted draft was the starting point for both. |
| Buying office or agent role | An AI tool drafts a summary sent to one buyer, built partly from another buyer's information seen the same day. | The agent who sends the summary is accountable for keeping each buyer's information separate, regardless of what the AI tool combined behind the scenes. |
| Different product families | Socks, knit-to-garment, terry and home-textiles businesses all assign accountability the same way; only the specific role and decision change with the product. | The rule that one named person is always accountable holds across every product family without exception. |
Suez Denim weaves Order B's 12 oz denim with 2% elastane for Delta Garments, working across a hybrid mix of order types for more than one buyer at once. A junior technical assistant uses an AI tool to help draft the fabric consumption figure for a new specification sheet, and the AI tool's draft understates Order B's own fixed consumption, showing 1.35 m per piece instead of the correct 1.45 m per piece.
The assistant sends the draft specification for release without noticing the discrepancy, because the figure looked reasonable and matched the format of past sheets. Karim's counterpart, Suez Denim's technical manager, is the person whose sign-off releases any specification sheet to a buyer or a garment factory; he is accountable for this error, not the assistant who used the AI tool and not the AI tool itself, because the sign-off step exists specifically to catch this kind of mistake before release.
When the error is found, the technical manager does not treat it as "the AI's mistake." He corrects the specification, records what happened and why the sign-off step did not catch it that time, and adjusts the review step so the same field is checked against the order's own fixed figures before every future release.
The worked numberUSD 960 is a real shortfall against Order B's own denim requirement if the wrong figure had been used to calculate how much fabric to weave and cut, on top of the disruption of a late fabric shortage discovered during cutting. The technical manager's sign-off step, exercised properly, is what catches this before it reaches Delta Garments' cutting room; when accountability is clearly assigned and actually exercised, the cost stops at the correction, not at a production stoppage.
That explanation feels complete because it names a cause, but a cause is not the same as accountability: an AI tool cannot correct itself, cannot be trained not to repeat the same mistake through discipline, and cannot answer to the buyer or factory that was affected. Treating the AI tool's error as the end of the conversation leaves the real question, whose sign-off should have caught it, unanswered.
The better rule is simple: name the accountable person before the AI tool is used, and hold that person to the same standard as if no AI tool had been involved.
An AI tool can help write, calculate or suggest, but it cannot be the person who answers when something goes wrong, because it cannot explain itself, cannot be corrected the way a person can, and cannot face the buyer or the factory that was affected. Before you let an AI tool help with a real decision, decide in your own mind who is the one person who checks its work and answers for it if it is wrong. That person does not change because an AI tool was involved.
When something does go wrong, fix the mistake and fix the review step that should have caught it, rather than stopping at "the AI tool made an error."
After the 0.10 m understatement is discovered, Karim contacts Suez Denim to ask how the wrong figure reached the specification sheet Delta Garments received. The technical assistant who drafted the specification with the AI tool tells Karim that "the AI tool got the number wrong," and that the assistant simply copied what it produced.
Should Karim accept "the AI tool got it wrong" as the full answer to what happened? Decide before opening the answer.
No. Karim asks the technical manager directly why the required check against Order B's own fixed denim consumption did not catch the error before sign-off.
1. The specification carried the technical manager's sign-off, which means the accountable person for this release is the technical manager, not the assistant who used the AI tool. 2. Suez Denim's own review step exists specifically to check this figure against the order's fixed specification, so the real question is why that step did not work this time, not what the AI tool produced. 3. Accepting "the AI got it wrong" as the full answer would leave the actual gap in the review step unfixed, which means the same kind of error could reach a future specification sheet unnoticed again.
The cost of accepting the incomplete answer is not just this one error; it is that the review step's own weakness would remain unaddressed, and the next AI-assisted error would have the same chance of reaching Delta Garments unchecked.
Order B's (the jeans order) fixed total order value shows what a specification error left unchecked could ultimately put at risk if it were not caught at the sign-off step:
The USD 960 denim fabric shortfall calculated earlier is a fraction of Order B's full value, but it is only that small because the sign-off step, once properly exercised, caught the error before it reached the cutting room. Had the same kind of error passed unnoticed into a larger commercial or construction field, rather than a consumption figure, the exposure could reach a much larger share of the order's own USD 29,400 total value. The cost of a clearly named and properly exercised accountability step is the time it takes to check a figure against the record; the cost of skipping it is whatever the uncaught error eventually reaches.
Use the T18-C05 AI governance and risk checklist workbook. In the Accountability sheet, list each kind of AI-assisted decision in the business, the accountable person who must sign off on it, and what should happen if an error is found. Start with Suez Denim's own specification sign-off on Order B's denim, then repeat the exercise for your own business.
[Download the AI governance and risk checklist workbook](../../templates/T18-C05-ai-governance-and-risk-checklist.xlsx)
1. Why is Suez Denim's technical manager accountable for the wrong fabric consumption figure, even though a junior assistant used the AI tool that produced it?
2. An export buying office uses an AI tool to draft a reply for Order A that mixes in a detail from another buyer's confirmed order by mistake, and the agent sends it without checking. Who is accountable for the mixed information reaching Northline Apparel?
| EN | AR | FR | TR | Meaning |
|---|---|---|---|---|
| Accountable person | الشخص المسؤول | Personne responsable | Sorumlu kişi | The one named individual who answers for a business decision, whether or not an AI tool helped produce it. |
| Attribution error | خطأ في تحديد المسؤولية | Erreur d'attribution | Sorumluluk atfı hatası | Blaming an AI tool for a mistake instead of naming the accountable person who used its output without checking it. |
| Sign-off | التوقيع بالاعتماد | Validation signée | Onay imzası | The accountable person's confirmation that an AI-assisted output has been checked and may be acted on. |
| Correction record | سجل التصحيح | Registre de correction | Düzeltme kaydı | A written note of what went wrong, who found it, and what was done to fix it, kept once an AI-assisted error is confirmed. |
| Escalation | تصعيد | Escalade | Yükseltme (eskalasyon) | Passing a decision or a found error up to a more senior or more accountable person when it is beyond the current person's authority to resolve alone. |
Next you will bring the use policy, the known risks, the adoption checks and this lesson's accountability rule together into one complete governance and risk checklist you can apply to any AI tool your own business already uses or is considering adopting. Back to the course map