# AI found the lowest quote. Why can’t we place the order yet?

> A fictional quotation exercise shows how to check sources, missing information and decision boundaries.

- URL: https://hyperfocusam.com/blog/checkable-ai-supplier-comparison/
- Language: en
- Author: Sam Wong (黃力桐), https://hyperfocusam.com/
- 中文版: https://hyperfocusam.com/blog/checkable-ai-supplier-comparison-tc/ (Markdown: https://hyperfocusam.com/blog/checkable-ai-supplier-comparison-tc/index.md)
- Published: 2026-09-23; revised 2026-09-23

---

You receive four quotations, give them to AI and ask which is cheapest. With the necessary information, that question can have a clear answer. But purchasing work rarely ends with identifying the lowest unit price.

That is a useful point to examine when evaluating an AI workshop. Does the exercise teach learners to check the answer and deal with the questions it leaves unresolved?

The packaging exercise below is adapted from a fictional case in existing teaching materials. Suppliers are labelled A to D. The table, instructions and reference answer have been reworked for this article. They are not an unedited AI conversation or advice about a real purchase.

## Start with the information available

Suppose you are comparing printed sleeve quotations for a food project. This extract includes unit prices, minimum order quantities, lead times and payment terms. It does not include freight, quotation validity or complete specification documents.

| ID | Unit price (HKD) | Minimum order | Lead time | Payment terms | Food-contact documentation | Sample |
|---|---:|---:|---|---|---|---|
| A | 0.88 | 10,000 | 14 days | 30 days | Table records it as provided | Ready |
| B | 0.78 | 8,000 | 12 days | Not stated | Available on request | Not provided in source |
| C | 0.95 | 6,000 | 9 days | 50% deposit; balance before delivery | Table records it as provided | Not ready |
| D | 1.08 | 5,000 | 18 days | 45 days | Table records it as provided | Ready |

The answer guide used for this adaptation does not state B’s sample status, so it remains unknown. A plausible value should not be filled in when the source does not supply it.

The fictional case also has purchasing rules. Payment terms must be clarified before selecting a supplier. A copy of the food-contact documentation must be obtained and checked. Deposits or payment before delivery require Finance review. These are exercise conditions, not Hong Kong law or rules for every organisation.

## The first answer completes only part of the job

B has the lowest listed unit price of the four. You can check that directly. But “B is cheapest, so choose B” goes further: it turns a price ranking into a purchasing decision.

Payment terms are not stated, and the documentation is only available on request. Even if everything eventually meets the requirements, the current information does not support that decision yet. Without freight and other details, the lowest unit price cannot be described as the lowest total cost either.

The useful next step is to separate what is known, what is missing and what needs to happen next.

## Turn the instruction into a checkable task

Try the following instruction with the downloadable table and rules. Use an organisation-approved tool; the exercise requires only fictional information.

```text
Using the attached quotation table and fictional purchasing rules,
prepare a supplier comparison.

Columns: Supplier ID | Explicitly stated information | Missing information |
Follow-up question | Source field.

Use only the attachments. Write “Not provided in source” for missing information.
“The table records documentation as provided” does not mean you have verified it.
Distinguish unit price from total cost. Do not select a supplier or commit to
an order or payment.
Finish by identifying which judgments cannot yet be made.
```

The instruction makes the checking requirements explicit. It does not guarantee that the tool will follow them. The result still needs to be checked against the original information, item by item.

## A usable answer should keep these gaps visible

The following reference has been prepared from the exercise data for this article. It is not the result of a fresh model test.

| Supplier | What can be confirmed | What still needs attention |
|---|---|---|
| A | Listed unit price is HKD 0.88; minimum order is 10,000 | Actual demand is not supplied, so the minimum order’s suitability cannot be judged. Documentation still needs verification. |
| B | Lowest listed unit price, HKD 0.78 | Ask for payment terms and a copy of the documentation. Sample status is unknown. |
| C | Shortest listed lead time, 9 days | Confirm when a sample will be available. Deposit and pre-delivery payment terms require review. |
| D | Listed payment term is 45 days; lead time is 18 days | Required arrival date is not supplied, so delivery suitability cannot be judged. |

If AI fills in B’s payment terms as “30 days,” that item fails the check, however polished the rest of the analysis looks. If it describes A as “fully compliant,” that also needs correcting. A statement in the quotation table and verification of a document are separate things.

An answer that preserves “Not provided in source” and asks the right follow-up question has done something useful.

## Then draft the enquiry

After checking the comparison, you could choose B as the recipient of an enquiry and ask AI to draft it using only the verified gaps. The message should ask for payment terms, a copy of the food-contact documentation and sample arrangements. It should not confirm supplier selection or authorise production.

Separating the analysis from the email lets you check what needs asking before checking whether the outgoing wording goes beyond that intention. The responsible colleague still reviews the email before sending it.

## How can HR/L&D assess this exercise?

Ask learners to submit their comparison and enquiry draft, then point out a source field, a missing item and a decision they cannot yet make. The trainer has something concrete to respond to, beyond asking whether everyone understood.

For a further check, change the dataset so that a different supplier has the lowest price and different fields are missing. See whether learners can identify the new problems independently. That tests their understanding of the checking method more directly than repeating the same prompt on the same data.

This exercise demonstrates how an AI task can be taught in a way that makes checking possible. It does not establish continued use at work or measure time savings. Those questions need separate evidence after the course.