Sourcing a new product usually means messaging a dozen-plus suppliers, most of whom reply in uneven English, at different speeds, with different terms buried in the wording. For a small importer without a dedicated sourcing team, that's a lot of manual reading and re-reading. AI is genuinely useful here for the reading and drafting, less so for the actual negotiating.
Step 1: Standardize your first-contact message
Before automating anything, write one solid first-contact template covering what you need to know from every supplier: MOQ, unit price at your target quantity, lead time, and payment terms. Sending a consistent message makes every reply easier to compare, and easier for AI to summarize consistently later.
Step 2: Let AI summarize and translate incoming replies
Supplier replies often mix broken English with pasted spec sheets. Feed each reply into ChatGPT or Claude and ask for a short, structured summary: MOQ, price, lead time, payment terms, anything unclear or missing. This is where the time actually gets saved: you're scanning a five-line summary instead of decoding a paragraph of dense supplier English for each of a dozen replies.
Step 3: Draft (don't auto-send) your follow-up questions
Once you can see which suppliers are worth a second round, use AI to draft your follow-up (pushing on price, asking for samples, clarifying payment terms). Read every draft before sending. Pricing and payment terms are exactly the place where a wrong assumption or a mistranslated number costs real money. This is not a step to fully automate.
Step 4: Automate the tracking, not the decisions
Zapier or Make can watch your inbox and log each supplier reply (summary, quoted price, lead time) into a shared sheet automatically, so nothing falls through the cracks across a dozen open threads. Make is worth it once you want branching logic, e.g. routing a reply that mentions a price above your target straight to a "needs negotiation" list instead of the default queue.
Where to stop automating
- Never let AI commit to a price, MOQ, or payment term on your behalf
- Don't skip reading a supplier's actual reply just because you have a summary; the summary can miss a caveat buried in the original wording
- Treat AI-drafted translations of contracts or purchase terms as a first pass only; anything binding should get a human (or professional translator) check
A first-contact template you can copy
Step 1 is easier with a template in hand. Keep it short and specific: suppliers answer clear, numbered questions more completely than a long introduction.
Subject: Enquiry: [PRODUCT NAME], [TARGET QUANTITY] units
Hello [SUPPLIER NAME],
I run [YOUR COMPANY], a small importer selling [PRODUCT CATEGORY] in [YOUR MARKET].
I'm looking for a supplier for [PRODUCT NAME, with key spec: material, size, finish].
Could you reply with:
1. Minimum order quantity (MOQ)
2. Unit price at [TARGET QUANTITY] units, and the price terms it is quoted on
3. Lead time from deposit to goods ready
4. Payment terms
5. Whether samples are available, and their cost
A spec sheet or photos of a similar item you've made would help.
Thank you,
[YOUR NAME]Asking what terms the price is quoted on is worth the extra line. Two quotes that look alike can include very different things, and you want to know before you compare them.
A prompt for summarising supplier replies
Use the same prompt for every reply so the summaries line up:
Summarise this supplier reply for a sourcing comparison.
Return exactly these fields:
Supplier | MOQ | Unit price (as quoted, with currency and quantity) | Price terms | Lead time | Payment terms | Samples | Unclear or missing | Caveats or conditions mentioned
Rules:
- Copy numbers exactly as written. Do not round, convert currencies or calculate totals.
- If a field isn't mentioned, write "not stated".
- List every condition attached to a price or lead time, even if it seems minor.
Reply:
[PASTE SUPPLIER REPLY]The "caveats" field is there because of the warning above: a summary can miss a condition buried in the original. Asking for it explicitly makes that less likely, though you still read the original for any supplier you're seriously considering.
Example: three replies for one product
Take a fictional importer of bamboo kitchen utensils as an example. She sends the template above to a shortlist of suppliers and gets three replies back in the first week, one with an attached price list and two in mixed English.
The summaries make the first comparison quick. Supplier A has the lowest unit price; Supplier B has the shortest lead time; Supplier C has a higher MOQ than she needs. On paper, A looks best.
Reading A's original reply before following up, she finds what the summary flattened: the low price applies only above a quantity well over her target, and it doesn't include the packaging she specified. She asks AI to draft a follow-up to A asking for the price at her actual quantity with packaging included, and a separate one to B asking about samples. She edits both, changing the AI's "we are ready to place an order soon" to "we're comparing a few suppliers", because she isn't.
That edit is the whole point of Step 3. A polite draft can still commit you to a position you haven't taken.
Common mistakes in AI-assisted sourcing
- Sending different first-contact messages to different suppliers. Replies then answer different questions and the summaries can't be compared.
- Comparing quotes on unit price alone. Price terms, MOQ, packaging and payment terms change what a price actually means.
- Letting AI convert currencies or units. Do conversions yourself, with a rate you choose, so you know where the figure came from.
- Follow-up drafts that sound more committed than you are. Phrases like "ready to order" or "happy with your price" read as signals to a supplier. Cut them unless you mean them.
- Logging summaries without the original. Keep a link to the original thread next to each row in your tracker, so you can check the wording later.
Checklist before you shortlist a supplier
- You have their MOQ, unit price at your quantity, price terms, lead time and payment terms in writing.
- You've read their original reply, not only the AI summary.
- Any translated figure or term you're relying on has been checked against the original text.
- You know what is and isn't included in the price (packaging, labelling, samples).
- Their row in your tracker has a status and a dated next action.
FAQ
Does this work for suppliers who don't reply in English at all? Yes, both ChatGPT and Claude handle translation reasonably well for sourcing-level correspondence, but treat the same caution as above: a translated number or term in a binding document deserves a second look before you act on it.
Is this worth setting up for just a few suppliers? Probably not. The summarize-and-log step earns its keep once you're juggling replies from eight or more suppliers at once. Below that, just reading each reply directly is faster than building the automation.
Can AI negotiate price with suppliers for me? It can draft your side of the negotiation, and it's useful for wording a counter-offer politely in plain English. It shouldn't decide the number or send the message. You set the price you're asking for; AI helps you say it clearly.
What columns should my supplier tracker have? At minimum: supplier name, status, next action, follow-up date, and the four key terms from your first-contact message (MOQ, unit price, lead time, payment terms), plus a link to the original thread. Notion vs a spreadsheet for follow-ups has a full column layout and a prompt that turns a supplier message into one tracker row.
Should I tell suppliers I'm using AI to translate? There's no need to announce it, but if a reply looks confusing, ask the supplier to confirm the key figures in their own words rather than guessing. A short "just to confirm, the price is X at Y units?" saves far more trouble than it costs.
What if suppliers reply on WhatsApp instead of email? The summarise-then-log approach still works; you paste the chat text in instead of an email. See automating WhatsApp Business support with AI for the WhatsApp-specific rules.
For reply summaries and drafting, see ChatGPT and Claude. For logging and routing replies automatically, see Zapier and Make. The same draft-then-approve principle applies to Shopify customer support automation. Read every AI draft before it goes to someone outside your team. If you're weighing Zapier against Make for the routing step, see Zapier vs Make for a two-person team; if your supplier tracker is outgrowing a spreadsheet, see Notion vs a spreadsheet for follow-ups.
