You Asked AI to Compare Two Options. Are Both Using the Same Facts?
A symmetrical comparison table can hide unequal evidence. Let missing information remain a question the buyer can follow up.

One supplier describes the box dimensions and price per pack. The other gives a unit price and a broad statement about material. The AI comparison places the two offers in a neat table and picks a favourite.
Before accepting that conclusion, check whether the table compares the same facts on both sides.
Choose the comparison questions first
In a fictional small shop, an operations buyer is reviewing packaging samples. She needs to know which option fits the product and the practical ordering task.
Write a shared set of questions before asking for a comparison: dimensions, quantity basis, included items and any other requirement that matters to this choice. Use the same units and the same scope where the source supports them.
This is a working comparison, not a claim that every purchase can be reduced to a score. A factor can be important even when it is not expressed as a number.
Keep missing information genuinely missing
If an offer does not state a relevant detail, mark it as not supplied. Do not let a table’s symmetry encourage the draft to complete every cell.
Separate a quoted fact from a calculation or assumption. If you convert a pack amount to a unit amount, record the quantity used and check the arithmetic yourself. A useful table makes that route visible.
Where the offers use different definitions, explain the mismatch. Two identical-looking labels can conceal different included quantities or conditions.
Connect each important entry to its source
Keep the source excerpt or reference beside the comparison entry. A reviewer should be able to find the detail without searching through an entire correspondence thread.
Ask the AI tool to extract and organise the supplied information before asking it to recommend. Review that extraction first. A recommendation built on an inaccurate table inherits the problem while making it less visible.
For the box example, a missing internal dimension may be a question for the supplier, not a reason to assume the product will fit.
Decide what can be concluded now
You may have enough information to reject an unsuitable size while still lacking enough to choose the final supplier. State that partial conclusion clearly.
Write the next questions needed to complete the comparison and identify who will obtain the answers. Keep the decision criteria visible so a new detail can be assessed consistently when it arrives.
Practise with fictional offers containing one missing field and one different quantity basis. Can a colleague distinguish facts, calculations and open questions in the finished table?
A fair comparison does not need to pretend that every option has supplied equal information. Its strength is showing which conclusions the evidence supports and which question comes next.
Continue with a related working question
Explore Turn Mixed Notes into Documents the Team Can Act On, or read The Summary Is Neat. Where Did the Unresolved Question Go?.
For a written learning option, see the AI at Work with Matze course overview. You provide an approved AI tool and perform the human review.
Review the provider’s course details. Affiliate link: Superstore may earn a commission if you buy.




Discussion 0
What did this article make you think about?