AI Supply Chain BriefChatGPT prompts for procurement: 12 prompts with worked examples
Short answerThe best procurement prompts give ChatGPT a role, the exact inputs, the output format and one rule: "list missing data instead of guessing." Use them for RFQ drafts, quote comparisons, negotiation plans, contract-term extraction and supplier emails. Paste only anonymized data, use your company's approved AI tool, and check every number before it leaves your desk.
Written by a working buyer. Every example on this page uses made-up data.
Most "ChatGPT prompts for procurement" lists give you 50 one-liners. This page gives you 12. Each one comes with when to use it, the full prompt, what good output looks like and what to check before anything leaves your desk. They also work in Microsoft Copilot, Gemini or any other assistant your company approves. For context, a 2025 Hackett Group release reported that about 49% of procurement teams had piloted Gen AI use cases and 4% reported large-scale deployment. Plenty of teams are still at the "try it on real work" stage, which is what these prompts are for.
How do you write a procurement prompt that works?
Give the AI five things: a role ("You are a senior buyer"), the exact inputs (pasted, anonymized), the steps you want, the output format (a table, a 150-word email) and one rule: "list missing data instead of guessing." Add a length cap so you get something usable, not an essay. Every prompt below follows that pattern.
Before you paste anything: a 4-point data-safety check
- Anonymize. Replace supplier names, part numbers and prices with codes like "Supplier 01".
- Use your company's approved tool. OpenAI says it doesn't train its models on business data by default (ChatGPT Business, Enterprise and the API). Consumer accounts work differently, and your company's policy decides what's allowed.
- Keep export-controlled and ITAR data out of general AI tools entirely.
- Check the math. Recompute totals in a spreadsheet. Researchers testing ChatGPT on spreadsheet problems found that its accuracy breaks down when information is limited or the problem is complex.
Sourcing and RFQs
1. RFQ draft from a spec
When to use it: You have a spec and quantities and need a clean RFQ package fast.
You are a senior buyer. Draft an RFQ for [ITEM] using the spec and quantities below. Include: scope, quantities and delivery schedule, quality and documentation requirements, the commercial terms I want quoted ([PAYMENT TERMS, INCOTERMS, PRICE VALIDITY]), a quote table suppliers must fill in (unit price, tooling, freight, lead time, MOQ, exceptions), and a response deadline of [DATE]. List anything missing from my inputs that a supplier would need to quote. Under 600 words. [PASTE ANONYMIZED SPEC AND QUANTITIES]
What good output looks like: A structured RFQ with a fill-in quote table and a short "missing information" list (for example, no packaging spec or no annual volume).
What to check: Every requirement traces back to your spec. Delete anything the AI added that you didn't ask for, such as certifications you don't need.
2. Bid comparison and gap finder
When to use it: Quotes are in and they aren't apples to apples.
You are a procurement analyst. Compare the anonymized quotes below in one table: unit price, one-time costs, freight, lead time, MOQ, payment terms, price validity and stated exceptions. Calculate total landed cost per unit at [ANNUAL VOLUME], spreading one-time costs over [YEARS]. Then list every gap or exception per supplier as a question I should ask. Do not pick a winner. Show your formulas. [PASTE QUOTES]
What good output looks like: A normalized table, landed cost per supplier with the formula shown, and a question list per supplier. The filled-in RFQ bid comparison example shows what this looks like.
What to check: Recompute landed cost in a spreadsheet, and make sure exceptions buried in the quote notes made it into the table.
3. Clarification questions to suppliers
When to use it: A quote is vague, or a supplier took exceptions you don't understand.
You are a buyer preparing supplier clarifications. From the quote below, write a numbered list of clarification questions covering price basis, what's included and excluded, lead time assumptions, MOQ, tooling ownership and any exceptions. Keep each question to one sentence and neutral in tone. Then draft a short cover email asking for answers by [DATE]. [PASTE ANONYMIZED QUOTE]
What good output looks like: Five to ten specific, answerable questions, not generic ones like "please confirm pricing".
What to check: No question reveals a competitor's price or your budget.
Negotiation prep
4. One-page negotiation plan (BATNA, walk-away, trades)
When to use it: You have a renewal or price talk coming and want to walk in prepared.
You are a negotiation coach for a buyer. Using the facts below, build a one-page plan: my objective, my BATNA, my walk-away point, the supplier's likely position and pressures, three things I can trade (volume, term length, payment terms, forecast visibility), my opening position with the reason for it, and the three hardest questions the supplier may ask, with answers. Use only my facts. Mark assumptions clearly. [PASTE ANONYMIZED FACTS: current price, volume, alternatives, timing]
What good output looks like: A plan that fits on one page, with your numbers in it and assumptions marked.
What to check: The walk-away point is yours, not the AI's suggestion. Confirm your authority limits before using it.
5. Price-increase pushback
When to use it: A supplier sent a price-increase letter.
You are a senior buyer. Here is a supplier's price-increase letter [PASTE]. 1) List each cost driver it claims and whether it gives evidence. 2) Draft a 150-word reply that acknowledges the letter, asks for a cost breakdown by driver with index evidence and dates, and says we will review once we have it. Don't accept or reject the increase. Don't invent figures.
What good output looks like: A list of claims marked "evidenced" or "not evidenced", plus a polite holding reply that doesn't commit to anything. The full method is in our guide on how to respond to a supplier price increase.
What to check: The reply doesn't commit to anything outside your authority. Check the contract's price-change terms separately.
6. Role-play the supplier
When to use it: You want to rehearse before a tough call.
Play the account manager of a supplier who wants a [X]% price increase on [ITEM]. Stay in character, push back realistically, and use common supplier arguments (raw materials, labor, capacity). I am the buyer. After 6 exchanges, step out of character and give me feedback: where I conceded too early, what I didn't ask, and one better line I could have used.
What good output looks like: Realistic pushback, then specific feedback tied to what you actually said.
What to check: Treat the feedback as practice, not a prediction of what your supplier will do.
Contracts and suppliers
7. Extract key contract terms
When to use it: You need the commercial terms from a long agreement quickly.
You are a contracts analyst helping a buyer. From the anonymized agreement below, extract into a table: payment terms, price adjustment or escalation terms, delivery and liquidated damages, warranty, limitation of liability, termination, and IP or tooling ownership. For each, quote the clause number and the exact wording. If a term isn't in the document, write "not found". Don't summarize beyond the table. [PASTE ANONYMIZED CONTRACT TEXT]
What good output looks like: A table with clause numbers and quoted wording, and honest "not found" rows.
What to check: Open every clause number and confirm the quote. This is a reading aid, not legal advice, so send real questions to your legal team.
8. Supplier onboarding checklist
When to use it: A new supplier is being set up and you want nothing missed.
You are a supplier quality and procurement lead. Build an onboarding checklist for a new [COMMODITY] supplier, grouped into: commercial (terms, pricing, banking setup), quality (certifications, first article, PPAP or equivalent if applicable), compliance (the declarations or screenings my company requires: [LIST]), logistics (packaging, labeling, Incoterms) and systems (vendor master, portal access). Add an owner column and a "done" column. Ask me for anything you need to tailor it.
What good output looks like: A grouped checklist with owners, tailored to the commodity.
What to check: Compare it to your company's own onboarding procedure. Where they differ, your procedure wins.
Everyday buyer work
9. Spend-line categorization
When to use it: You have uncategorized AP or PO lines and need them grouped. This is the worked example below.
You are a procurement analyst. Categorize each line below into exactly one of: Fasteners, Packaging, MRO/Safety, Freight, Services – Calibration, IT Hardware, Unknown. Add a confidence column (High/Med/Low) and a short note. Do not create new categories. Mark Low for anything ambiguous and say what information would resolve it. Then total the amounts by category. [PASTE ANONYMIZED LINES: line, supplier code, description, amount]
What good output looks like: Every line in exactly one category, honest Low-confidence flags, and category totals.
What to check: Category totals add up to the export total, and nothing is forced into a category it doesn't fit.
10. Messy email thread to status update
When to use it: Your manager asks "where are we with that order?" and the answer is buried in 12 emails.
Summarize this email thread for my manager in 5 bullets: current status, dates promised, open decisions, who owns each, and risks. Then list the three questions I still need answered. Use only what's in the thread. If a date or quantity is unclear, write "unclear". [PASTE ANONYMIZED THREAD]
What good output looks like: Five factual bullets and three open questions, with "unclear" wherever the thread is.
What to check: Check every date and quantity against the original emails.
11. Late-PO follow-up email
When to use it: A supplier has missed a confirmation or a ship date.
Draft a firm, polite follow-up email to a supplier about PO [NUMBER] for [QTY] [ITEM], originally due [DATE]. Ask for a confirmed ship date, the reason for the delay, and a recovery plan if the date can't be met. Ask for a reply by [DATE/TIME]. Under 120 words. No threats, no legal language.
What good output looks like: A short email with three clear asks and a reply-by time.
What to check: The PO number, quantity and dates are right, and the tone fits your relationship with this supplier.
12. Weighted supplier scorecard
When to use it: You need to compare suppliers on more than price.
You are a sourcing analyst. Build a weighted scorecard for comparing [N] suppliers of [ITEM]. Criteria and weights: [e.g., landed cost 40%, quality 25%, delivery 20%, risk 15%]. Define what a 1, 3 and 5 score means for each criterion so two people would score the same way. Then score the suppliers from the facts below, show the weighted totals, and list which scores are judgment calls. [PASTE ANONYMIZED FACTS]
What good output looks like: Clear scoring definitions, a weighted total per supplier, and the judgment calls named.
What to check: The weights add up to 100% and the scores match your evidence. The award decision stays with you.
Worked example: categorizing 8 spend lines
Fictional exampleExample Industrial Supply Co., its suppliers and every number in this example are made up for illustration. Don't paste real supplier quotes, NDA pricing, personal data or export-controlled data into an AI tool. Anonymize first and follow your company's AI policy.
Example Industrial Supply Co. exports 8 uncategorized AP lines and runs prompt 9 on them.
The input
| Line | Supplier (anonymized) | Description | Amount |
|---|---|---|---|
| 1 | Supplier 01 | M8 x 25 hex bolts, zinc, box 500 | $212.40 |
| 2 | Supplier 02 | LTL freight, PO 4471 | $388.00 |
| 3 | Supplier 03 | Nitrile gloves, L, case | $96.00 |
| 4 | Supplier 04 | Annual calibration, torque wrenches | $1,150.00 |
| 5 | Supplier 05 | Corrugated cartons 18x12x10 | $2,460.00 |
| 6 | Supplier 06 | "Misc. services – Q3" | $4,800.00 |
| 7 | Supplier 01 | Flat washers M8 | $41.10 |
| 8 | Supplier 07 | Laptop docking station | $189.99 |
| Total | $9,337.49 |
The prompt
You are a procurement analyst. Categorize each line below into exactly one of: Fasteners, Packaging, MRO/Safety, Freight, Services – Calibration, IT Hardware, Unknown. Add a confidence column (High/Med/Low) and a short note. Do not create new categories. Mark Low for anything ambiguous and say what information would resolve it. Then total the amounts by category. [PASTE ANONYMIZED LINES: line, supplier code, description, amount]
What good output looks like
| Line | Category | Confidence | Note |
|---|---|---|---|
| 1 | Fasteners | High | – |
| 2 | Freight | High | – |
| 3 | MRO/Safety | High | – |
| 4 | Services – Calibration | High | – |
| 5 | Packaging | High | – |
| 6 | Unknown | Low | Ask the requester what service was provided and get the PO. |
| 7 | Fasteners | High | – |
| 8 | IT Hardware | High | – |
| Category | Total |
|---|---|
| Unknown | $4,800.00 |
| Packaging | $2,460.00 |
| Services – Calibration | $1,150.00 |
| Freight | $388.00 |
| Fasteners | $253.50 |
| IT Hardware | $189.99 |
| MRO/Safety | $96.00 |
| All lines | $9,337.49 |
The teaching point: the most useful line in the output is the one the AI refused to guess. Line 6, "Misc. services – Q3", is also the largest line at $4,800.00. A good prompt makes the AI say "I don't know" and tell you what would resolve it, here the requester and the PO.
What to check: the category totals add up to the export total of $9,337.49, and no line was forced into a category it doesn't fit.
What is ChatGPT bad at in procurement?
- Arithmetic on long tables. It can drop rows or slip a decimal. Recompute in a spreadsheet (see "ChatGPT and Excel: trust, but verify").
- Live prices and market data. Unless it's reading a source you can check, it doesn't know today's resin or freight price.
- Facts about specific suppliers. It can state wrong details confidently. Verify anything you'd rely on.
- The decision. Awards, price commitments and compliance calls stay with you.
For real company examples of AI in sourcing, see the RFQ and sourcing events, contract review and spend classification use cases. Planners can use the ChatGPT prompts for inventory management. If your company runs on Microsoft, see the Copilot prompts for procurement.
FAQ
How do I use ChatGPT for procurement?
Start with structured writing and comparison work: RFQ drafts, quote comparison tables, negotiation plans, contract-term extraction and supplier emails. Give it a role, anonymized inputs, the output format and a rule to flag missing data. Then check the output, especially any math, before you use it.
What are the best ChatGPT prompts for procurement?
The ones that define their inputs and output. The 12 on this page cover RFQs, bid comparison, supplier clarifications, negotiation prep, price-increase replies, contract terms, onboarding, spend categorization, status updates, late-PO follow-ups and scorecards. Each one tells the AI to list missing data instead of guessing.
Is it safe to put supplier quotes or contracts into ChatGPT?
It depends on the account and your company's policy. OpenAI says it doesn't train its models on business data by default for ChatGPT Business, Enterprise and the API, but consumer accounts work differently. Anonymize supplier names and prices, use your approved tool, and never paste export-controlled or ITAR data.
Can ChatGPT compare supplier proposals?
Yes, as a first pass. It can put the quotes into one table, calculate landed cost and turn gaps and exceptions into questions. It shouldn't pick the winner, and you should recompute the numbers in a spreadsheet. Prompt 2 and the RFQ bid comparison example show how.
What do buyers actually use AI for day to day?
Common tasks are drafting emails and RFQs, summarizing long threads, comparing quotes, preparing for negotiations and pulling terms out of contracts. These are writing and organizing jobs where you can check the output. Decisions and commitments stay with the buyer.
Should I use Copilot or ChatGPT for procurement?
Use whichever your company approves. These prompts work in both because they spell out the role, inputs and format. If your company runs Microsoft 365, Copilot works inside Outlook, Word, Excel and Teams. See our Copilot prompts for procurement guide.
