Can ChatGPT or Claude compare your vendor quotes?

Checked against public product pages on Oct 13, 2026. AI products change often. Test on your own quotes.

What can a general AI assistant do with a vendor quote?

It can do the reading and a first pass of the matching. It reads almost any format, needs no setup, and suits a one-off quote, an odd supplier, or a second check on a sheet someone already typed.

How do you set it up so the matches are accurate?

Make your list the authority, separate reading from matching, and forbid guessing.

1. Your material list is the rows. Export it to Excel or CSV with code, description, quantity and unit.
2. One vendor per conversation, so lines from different suppliers cannot mix.
3. Split long quotes. Anthropic's PDF documentation says dense PDFs with complex tables can fill the model's context before the page limit. Send a long waterworks quote one section at a time.
4. Extract first, match second. You can check the extraction against the PDF on its own.
5. A page and line for every price. That is your receipt.

A prompt you can use today

I am leveling supplier quotes for a bid. Accuracy matters more than speed.

Inputs:
1. MY MATERIAL LIST (attached). My rows are the authority. Columns: code, description, qty, unit.
2. ONE VENDOR QUOTE (attached PDF) from [vendor], quote number [ ], dated [ ].

Work in two passes.

PASS 1, EXTRACT ONLY. List every line on the quote exactly as written:
page | line or item no. | vendor description | qty | unit | unit price | extended price.
Include freight, tax, surcharges, price-hold dates, notes, exclusions and alternates as their own lines.
At the end, give the quote's printed subtotal and the sum of the extended prices you listed.

PASS 2, MATCH. For each row of MY list, return one row:
my code | my description | my qty | my unit | vendor page/line | vendor description | vendor qty | vendor unit | unit price | status | note

Rules:
- Status is one of: MATCHED, PARTIAL, NO QUOTE, NEEDS REVIEW.
- PARTIAL means size, class, pressure rating, joint type or material differs. Say what differs.
- Never guess. If you are not sure it is the same item, use NEEDS REVIEW and say why.
- Do not convert units unless the quote states the basis (for example, joint length). Show every conversion in the note.
- Do not change my rows, codes or quantities.
- After the table, list every vendor line you did not match to any of my rows.

Paste the result into your comparison sheet, not straight into the estimate.

Where do ChatGPT and Claude go wrong on supplier quotes?

They fail quietly. The output looks finished, so the errors are easy to miss.

• Forced matches. Asked to fill every row, a model picks the nearest item. An 8-in MJ gate valve and an 8-in flanged gate valve read alike. So do C900 DR18 and DR14. The price belongs to a different item.
• Units. Pipe priced per foot or per joint, aggregate per ton when you bid cubic yards. A model may convert without telling you, or not convert at all.
• Missing lines. A quote that leaves out joint restraints or accessory kits does not say so. A model that fills every row hides the scope gap.
• Long and scanned quotes. Section subtotals, scanned faxes and handwritten notes are where lines drop or get misread.
• Arithmetic. Check extended prices and totals in Excel, not in the chat.
• No receipt and no memory. The answer is a chat message, not attached to the bid. Each conversation starts over. You can attach an alias list of each vendor's descriptions every time, but you maintain it.
• Your data. Quotes carry vendor pricing. Check your plan's data settings and company policy before you upload.

How do you check the output before a price carries?

Reconcile it, then sample it against the PDF.

1. Count. Every row of your list appears once, with a price or a status.
2. Reconcile. The extracted extended prices add up to the quote's printed subtotal. If they do not, a line is missing or misread.
3. Sample. Open the PDF at the cited page for every PARTIAL and NEEDS REVIEW row and a sample of MATCHED rows.
4. Units. Every conversion is written in a note.
5. Gaps. Every NO QUOTE is confirmed as not quoted, not missed.

For the full method of leveling once the prices are clean, see How should you compare material vendor quotes?

ChatGPT or dedicated software: which should you trust?

Trust the workflow that leaves a receipt for every carried price and a person who confirmed it. Dedicated tools build that receipt in. With a general assistant, you build it yourself.

• HCSS HeavyBid. In its Summer 2026 product release (July 2026), HCSS says HCSS AI pulls pricing from an uploaded vendor quote and maps it to resource line items in the quote folder, and the estimator reviews, confirms or adjusts each suggested price before it is added. Its release webinar (recorded July 16, 2026) also showed the source page, a confidence rating and a warning when quantities differ.
• ProcurePro BidLevel extracts prices, inclusions and exclusions and traces every value to the original PDF.
• Field Materials matches the lines of a quote PDF to the material list in your RFQ and can learn your vendors' names for your items.

A general assistant is a good choice for a one-off quote or a second check. When the same work repeats on every bid, the prompt, the checks and the alias list become a job someone owns.

Is there a tool or AI agent that takes a material list and a quote and returns matched prices?

Yes. Each option above does a version of it, and they differ by who does the work: you with an assistant, your estimator inside HeavyBid, a buyer in a procurement platform, or a service such as Coffer, where you send the list and the quotes and the mapped, leveled comparison comes back. For a dated side-by-side, see What tools compare supplier quotes for heavy civil contractors in 2026?

What does a service add over doing it yourself?

A service takes the reading, matching and retyping off your desk. You keep the decisions.

Coffer is a quote-to-PO service for heavy-civil contractors. You send the material list and the vendor quotes that came back. Coffer reads, matches and levels every quote: it maps each line to your material list or HeavyBid resource codes, keeps gaps and unit differences visible, and returns the comparison sheet in your workbook with PO drafts for the vendors you pick. Coffer ties every proposed match back to its line on the source quote. It recommends a carry for each row and shows why. Your estimator makes the final carry decision. Each vendor's matches stay on file and inform the next quote from that vendor.

A $300M self-perform civil contractor in Texas is moving into deployment after seeing it on their own bids.

What is the next step?

Send one past bid's material list and the supplier quotes that came back to declan@coffer.build, or use the contact form below. You get the mapped, leveled comparison back to check against what your team carried.