Most procurement people use AI the same way they used Google in 2004: type a question, read the first result, move on. The output is fine. It is not impressive. It does not save significant time.

The difference between a mediocre AI interaction and a genuinely useful one is not which model you use. It is how you frame the question.

Three principles, then four procurement-specific examples you can use today.

Three principles before the examples

Assign a role. AI models calibrate their response to whoever they think they are talking to. "Review this contract" gets a generic response. "You are a procurement lawyer reviewing a supplier contract for a mid-size industrial manufacturer. Identify risks that are unusual or missing protections that you would expect to see in a standard manufacturing supply agreement" gets something useful.

Give context, not just a task. The more specific the context, the more precise the output. Industry, category, company size, your position in the negotiation — all of this shapes what a useful answer looks like. Generic context produces generic answers.

Specify the output format. If you want a table, say you want a table. If you want a numbered list with severity ratings, say that. AI will fill the space you give it. If you leave the format open, it will write paragraphs when you wanted a checklist.

Use case 1: Contract risk review

You have a supplier contract. You want to know what to pay attention to before signing or before the renewal negotiation.

Weak prompt
Review this contract and tell me if there are any risks.
Strong prompt
You are a procurement lawyer reviewing a supply agreement for a manufacturing company. The contract is with an external logistics provider for EMEA distribution. Review the following contract text and identify: 1. Automatic renewal clauses — exact wording and notice period required to cancel 2. Liability caps — are they capped, and at what level relative to contract value 3. Price escalation provisions — what triggers them, any caps 4. Exclusivity terms — any restrictions on using alternative suppliers 5. SLA definitions and penalty triggers — what specifically constitutes a breach 6. Termination for convenience — which party has it and under what conditions For each, rate the risk as Low / Medium / High and give one sentence on what to negotiate. [PASTE CONTRACT TEXT HERE]

The second prompt defines a role, specifies the six exact items to check, asks for a severity rating, and requests a one-sentence action. The first prompt asks the AI to decide what matters — and it will guess conservatively.

Use case 2: Spend data summary

You have exported spend data. You need to present it to your manager or use it in a negotiation. Manually writing the commentary takes an hour. AI can do it in 30 seconds if you prompt correctly.

Weak prompt
Analyse this spend data and tell me what you see.
Strong prompt
You are a procurement category manager preparing a quarterly spend review for indirect materials at a European manufacturing company. Analyse the following spend data (CSV format) and produce: 1. Total spend by top 10 suppliers — table format, sorted descending 2. Spend concentration: what % of total spend goes to the top 3 suppliers 3. Any supplier that appears multiple times under different names (likely duplicates) 4. Month-on-month trend: is spend increasing, decreasing, or flat 5. Two observations that would be worth raising in a negotiation with the top supplier Keep the summary under 300 words. Use plain language — this is for a manager who does not work in procurement. [PASTE SPEND DATA HERE]

The second prompt specifies the exact outputs (5 items), the format for each, the audience, and the word limit. You will get a usable first draft in one attempt instead of three rounds of refinement.

Use case 3: Supplier quote comparison

You have received three quotes for the same requirement. The formats are different, the payment terms are different, and one quote bundled in items the others quoted separately. Comparing them manually is slow and error-prone.

Weak prompt
Compare these three supplier quotes and tell me which is best.
Strong prompt
You are a procurement analyst comparing three supplier quotes for [CATEGORY]. The quotes are pasted below. They use different formats and may include different line items. Your task: 1. Normalise the quotes into a single comparison table. Columns: Supplier / Item / Unit price / Total / Payment terms / Delivery time / Warranty 2. Where a supplier has not quoted a specific item, mark it as "not quoted" 3. Calculate the total landed cost for each supplier assuming Net 30 payment terms and delivery to [LOCATION] 4. Identify any non-price differences that are worth flagging (lead time gaps, warranty differences, bundled items) 5. Recommend which supplier offers the best value — with one sentence justification Do not make assumptions about pricing. If something is unclear, flag it as a clarification needed. QUOTE 1: [PASTE] QUOTE 2: [PASTE] QUOTE 3: [PASTE]

The instruction "do not make assumptions — flag clarification needed" is important. Without it, AI will invent plausible numbers when something is ambiguous. That will look fine in the table and be wrong in the negotiation.

Practical note on confidentiality
Do not paste supplier contracts, pricing, or commercially sensitive data into consumer AI tools (ChatGPT.com, Claude.ai) without understanding your company's data policy. Use an enterprise-licensed tool with data privacy guarantees, or anonymise sensitive values before pasting.

Use case 4: Negotiation preparation brief

You have a negotiation coming up with a major supplier. You want to prepare arguments, anticipate their responses, and identify your walk-away point. This takes two hours to do well manually. With a good prompt, you get a draft in ten minutes.

Weak prompt
Help me prepare for a negotiation with my supplier.
Strong prompt
You are a senior procurement manager preparing for a price negotiation. Context: - Supplier: [NAME], provides [CATEGORY] under a framework agreement expiring [DATE] - Current annual spend: [VALUE] - Our leverage: we represent approximately [X%] of their revenue in this region / we have identified two qualified alternatives / our volume will increase by [X%] next year - Their likely position: they will argue input cost increases (materials, energy) justify a price rise of [X%] - Market context: [e.g. raw material index for steel is down 12% YTD / logistics rates have normalised since 2023 peak] Produce: 1. Our opening position and the rationale (2-3 sentences) 2. Three concessions we can offer that cost us little but matter to the supplier (e.g. longer contract, faster payment, volume commitment) 3. Three counter-arguments to their expected "input costs have risen" argument, with supporting logic 4. Our walk-away threshold and what we do if we reach it 5. One non-obvious question to ask early in the meeting to understand their actual cost pressure

The context section — especially "their likely position" and "market context" — is what makes this useful. Without it, you get generic negotiation advice. With it, you get something that resembles actual preparation.

The pattern across all four

In each case, the strong prompt does the same four things: assigns a role, provides specific context, defines the exact outputs, and specifies any constraints (word limit, format, what to do with ambiguity).

This is not a special skill. It is just structured thinking applied to a new tool. The procurement people who are doing this well are not AI enthusiasts. They are people who are organised enough to write a clear brief — and applied that same habit to their prompts.

Start with one of the four cases above. Copy the strong prompt, fill in your context, paste your document. See what comes back.