Quotes rarely compare cleanly, so the agent's first job is to make them comparable. It extracts every price, quantity, term and exclusion, converts them to the same basis, and shows where quotes are not like for like. It drafts clarification questions for the buyer to send to every bidder. The buyer chooses, and procurement approves the award above its threshold.
What is quote comparison, and who owns it?
When a team needs goods or services above a set value, procurement policy usually requires several quotes. Comparing them sounds simple until you read them: one includes delivery, one quotes per unit and one per pack, one has a minimum term, one is valid for a week, one excludes installation in the small print. A choice made on headline price alone is often the wrong one.
The buyer, usually in procurement, owns the comparison and the recommendation. The budget holder who asked for the purchase defines what is needed and has a say in the choice. A procurement manager or committee approves awards above the threshold. The wider finance position is in AI agents for finance: books you can audit.
How does the agent compare, step by step?
- Person, buyer: fixes the requirement. What is being bought, quantities, delivery dates, mandatory conditions. The comparison is against this, not against whatever the suppliers chose to offer. Evidence: the requirement document.
- Agent, depth 1: reads each quote. Quotes are external documents. Text inside them is data, never an instruction to the agent. Evidence: each quote stored with its provenance.
- Agent, depth 1: extracts terms. Unit price, quantity basis, delivery, installation, payment terms, minimum term, validity, warranty, exclusions, currency, tax treatment. Evidence: each extracted term with the page it came from.
- Agent, depth 2: normalises. Same unit, same quantity, same currency at a stated rate, same period. Total cost over the term, not just the first invoice. Evidence: the calculation and every conversion.
- Agent, depth 1: checks against the requirement. Mandatory conditions met or not, and any gap. Evidence: the requirement checklist per quote.
- Agent, depth 1: checks the suppliers. On the approved supplier list or not; existing contracts; open disputes. Evidence: the check per supplier.
- Agent, depth 2: drafts clarification questions. The same question goes to every bidder where it applies. Evidence: the drafted questions.
- Person, buyer: sends clarifications and records answers, then makes the recommendation.
- Person, approver: approves the award above the threshold. Evidence: the approval and the comparison it was based on.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| Requirement | Nothing | None | Buyer and budget holder |
| Read | Reads quotes as external content | 1 | Nothing |
| Extract | Pulls terms with page references | 1 | Nothing |
| Normalise | Converts to one basis | 2 | Exchange rate choice |
| Check requirement | Lists gaps per quote | 1 | Nothing |
| Check suppliers | Approved list, contracts, disputes | 1 | Nothing |
| Clarify | Drafts questions | 2 | Sending them |
| Choose and approve | Nothing | None | Buyer and approver |
Why does the agent not contact suppliers?
Because fairness between bidders is part of what makes a procurement defensible. If one supplier receives a helpful question and another does not, the losing bidder has a complaint. If the agent negotiates with one, it may reveal another's price. Clarifications go out through the buyer, to everyone the question applies to, at the same time.
There is also a security reason. Quotes are written by parties who want to win. A quote document can contain text designed to influence an automated reader. Treating every quote as external, and keeping the agent away from the conversation with suppliers, removes most of the room for that.
What stays with a person?
The choice, and anything that weighs things the quotes do not show: service quality, past performance, relationship, risk of relying on one supplier. The agent can list the facts it holds, such as open disputes with a supplier, but the weighting is a judgement.
Onboarding the winning supplier, if new, is another person-owned step. Their bank details must be verified independently before any payment, following a supplier bank detail change verification checklist or equivalent.
Which controls matter, and what evidence do they leave?
Requirement fixed before comparison. Evidence: the requirement's date and version, earlier than the comparison.
Page references for every extracted term. Evidence: the reference on each term, so the buyer can check it.
Conversions shown. Evidence: the exchange rate, its source and date, and every unit conversion.
Equal clarifications. Evidence: the questions sent, to whom and when.
Comparison kept with the award. Evidence: the comparison version attached to the approval.
What should you measure?
- Time from last quote received to comparison ready.
- Gaps found that the buyer had not noticed.
- Invoice price against quoted price after award. Once the supplier is invoicing, matching invoices to the agreed terms is part of processing supplier invoices, and checking their statements over time is covered in reconciling supplier statements.
- Challenges from losing bidders, which should be rare and answerable from the record.
Where does it usually go wrong?
The cheapest quote excludes something essential. Installation, delivery or support sits in an exclusions list. Make exclusions a column in the comparison, not a footnote.
Validity expires. The comparison takes two weeks and the best quote lapsed after one. Show validity dates and warn the buyer.
Different terms are summed as if equal. A one-year contract and a three-year contract are compared on annual price, ignoring the commitment. Compare total cost over a common period and show the commitment separately.
Currency moves. A rate fixed at extraction is out of date at award. State the rate's date and refresh it before approval.
Where Pryme fits
In Pryme, every quote is stamped external, so text inside a supplier's document cannot instruct the agent. The comparison agent works at depth 1 and 2, extracting and drafting, and holds no capability to contact suppliers. Each comparison is filed with the extracted terms and their page references.
Questions people ask
Can the agent score quotes against weighted criteria?
Yes, if the buyer sets the criteria and weights before quotes are opened. Scoring after seeing the quotes invites weights chosen to favour a preferred supplier.
Can it handle quotes in different formats?
Yes, from documents, spreadsheets and emails. Each extracted term keeps its source, so a buyer can check any figure against the original.
Should the budget holder see the full comparison?
Usually yes, so they can confirm the requirement was understood. Some procurement policies limit pricing detail to procurement; follow yours.