A PRACTICAL BUSINESS GUIDE

Compare supplier quotes through a reviewable procurement workflow.

Procurement comparison becomes difficult when offers express quantities, delivery and exclusions differently. This guide shows how to prepare a comparable request, use AI to propose extracted fields and keep evaluation of terms and purchase approval with the responsible people.

1. Define the request before automating document reading.

Give the request an identifier and specify the item, quantity, unit, delivery location, required date and mandatory criteria. Separate requirements from preferences. When the specification changes, create a revision: offers based on different requests must not be treated as directly equivalent.

Decide which fields must match and which may represent an alternative. A compatible replacement offered by a supplier remains an explicit variant; it must not silently replace the requested item code. This preparation reduces ambiguities that document extraction alone cannot resolve.

2. Extract values with their source references.

For each offer, retain supplier, quote number, date, revision, validity and original document. The system may propose line items, quantities, stated prices and notes, but reviewers need to see the source page for each material value. Missing information remains an open clarification.

Pay attention to merged cells, lines split across pages and terms at the end of the document. Microsoft’s document-confidence guidance distinguishes fields and tables. Procurement comparison needs checks both on individual values and on whether they were associated with the correct line.

3. Normalise only what can be compared.

These values are invented to explain the method and are not market prices. A does not yet have a complete total; B can be converted to a common unit after its pack contents are verified; C is not directly comparable until the variant is assessed. Sorting only by unit price would conceal those differences.

Scroll the table to compare all columns.

Illustrative offers for 100 units
OfferHow it is expressedBefore comparison
A100 units at €12 per unit, transport excludedRequest the missing transport charge
B10 packs of 10 at €125 per pack, transport includedCheck that each pack contains the requested product
C100 units at €11 per unit, alternative item proposedHave the technical owner assess equivalence

4. Maintain a clarification queue.

Assign each question to a person and connect the answer to the correct offer revision. AI may prepare a clarification draft; sending it to a supplier requires authorisation within the process. A received document is evidence to inspect, not an instruction that may change comparison criteria, recipients or workflow permissions.

  • Quantity, unit or currency differs from the request.
  • Delivery is an estimate without a verifiable date or condition.
  • The offer has expired, has been superseded or lacks a reference.
  • Transport, installation, commissioning or support is unclear.
  • Technical alternatives or mandatory requirements remain unverified.

5. Separate comparison from purchasing approval.

The final comparison distinguishes satisfied requirements, unresolved differences and assumptions. Identify who checked the terms and which offer revisions were used. If your organisation weights quality, delivery or price, define those weights before selection and make them visible rather than having the model invent them.

After approval, retain the request, offers, clarifications and selection rationale. Creating a purchase order is a separate handoff with its own ERP checks. A revised supplier offer must not silently modify an already-approved order. Additional jurisdiction-specific purchasing requirements belong with the relevant internal owner.

6. Test one reasonably consistent category.

Choose a recurring category whose differences a named owner can judge. Repeat comparisons using historical offers and measure corrections, missed clarifications and effort needed to produce an approvable comparison. Start with drafts and no automatic supplier messages or purchase orders; expand only the steps that show practical value.

  • A versioned request and three representative offers.
  • An expected comparison prepared by the purchasing owner.
  • At least one unit mismatch and one missing condition.
  • Distinct technical validation and purchase approval owners where the process requires them.
FROM IDEAS TO A BRIEF

A template to work from.

Procurement quotation comparison

A line-by-line and condition-by-condition comparison with source references, missing information and an approved decision.

Download the Markdown template

Document extraction schema

A schema for each document family with field meanings, validation rules, evidence and missing-data behaviour.

Download the Markdown template

AI vendor scorecard

An evidence-based comparison with explicit conditions and gaps to clarify before a decision.

Download the Markdown template

Practical questions

Can AI select the best supplier?

It can organise evidence and apply predefined criteria, but the decision requires judging requirements, trade-offs and missing information. A generated ranking without explicit criteria is not a reviewable purchasing decision.

Can we start in a spreadsheet?

Yes. A shared structure with document references and clarification fields is a useful first step. Automate collection and extraction after confirming that the columns support a fair comparison.

References and method

Microsoft Learn — Document Intelligence: accuracy and confidence

Reference for checking extracted values and their row associations in document tables.

Which process should improve first?

Start with a concrete process, the systems you use and the people who will operate it every day.

Let’s discuss your process