Most export companies still handle enquiries the same way: someone opens the inbox, reads each email and types out each reply by hand. In peak season there's no keeping up, new staff take a long time to get up to speed, and the style of each reply depends entirely on whoever wrote it.
Building an "enquiry automation" pipeline with AI splits the process into three steps: AI drafts the reply → a person checks it → it goes out automatically. You keep a human in control, while handing the most time-consuming part, writing from scratch, to the AI.
Step 1: Sort your enquiries
Enquiries arrive through different channels (your website form, the Alibaba.com back office, email). First gather them in one place. The simplest way is an automation tool like Zapier, which can sync new enquiries from each channel into one spreadsheet or one shared inbox so they can all be handled the same way.
Step 2: Have AI draft the reply
Feed the enquiry details (the product model the customer is asking about, their target price, their order quantity) into ChatGPT or Claude and have it draft a reply that follows your company's pricing rules and tone. Both handle routine enquiries well. If the customer's question involves quoting from a long product specification or contract terms, Claude is usually a little more reliable with long documents.
A prompt can be as simple as this:
You are an export sales representative for our company. Your tone is professional and concise. Based on the customer enquiry below, draft a quotation reply email. Our pricing rules are: [PRICING RULES]. Original enquiry: [PASTE ENQUIRY]
Step 3: A person checks it before it goes out
Every AI draft must be reviewed by a salesperson before it's sent, especially anything involving specific prices or delivery commitments: the AI may misread what the customer wants, or quote outdated pricing rules. Make "AI drafts, a person approves" a fixed routine for the team, rather than letting the AI send emails on its own.
Going further: connect the whole workflow with an automation tool
If you get a lot of enquiries, you can use a tool like Zapier to chain "new email arrives → call the AI API to generate a draft → write it into a review spreadsheet → notify the salesperson" into one automated workflow. The salesperson then only needs to open the spreadsheet, approve the draft, copy it and send it, without switching between channels by hand. If the workflow needs to branch depending on where the enquiry came from or how large the order is, see Zapier vs Make for a two-person team.
Price, MOQ and certification: prompt templates for three types of enquiry
If you prepare a separate prompt for each type of enquiry, the AI's drafts will be much closer to how your business really works. Below are templates for the three most common types; replace the parts in square brackets with your own company's rules. For a general template for everyday quotation emails, see ChatGPT vs Claude for B2B trade emails.
Price enquiries
You are an export sales representative for [COMPANY NAME]. Your tone is professional and concise, and you reply in the language the customer wrote in.
Using the pricing rules below, draft a quotation reply to this enquiry.
Pricing rules: [PRICE FOR EACH MODEL, TIERED PRICING BY QUANTITY, INCOTERMS, QUOTE VALIDITY]
Requirements:
- Prices may only come from the pricing rules. Do not quote for any model that is not in the rules; write "We will confirm this model and send you a separate quote" instead
- Ask for any information the customer left out: [FOR EXAMPLE: QUANTITY, PORT OF DESTINATION, WHETHER A CUSTOM LOGO IS NEEDED]
- Do not give a specific delivery date; write [TO BE CONFIRMED] instead
Original enquiry: [PASTE THE ENQUIRY]MOQ enquiries
You are an export sales representative for [COMPANY NAME]. The customer is asking about our minimum order quantity.
Our MOQ rules: [MOQ FOR STANDARD ITEMS, MOQ FOR CUSTOM ITEMS, WHETHER MIXED ORDERS ARE ALLOWED]
Please draft a reply that states the MOQ. If the customer needs less than the MOQ, offer only the alternatives the rules allow (for example a mixed order, or switching to a standard item). Do not mention any option that is not in the rules.
Original enquiry: [PASTE THE ENQUIRY]Certification enquiries
You are an export sales representative for [COMPANY NAME]. The customer is asking about product certifications or test reports.
Certifications and reports we hold: [CERTIFICATE NAME, PRODUCTS COVERED, VALID UNTIL]
Please draft a reply that mentions only the certifications listed above. If the customer asks about a certification we don't hold, say honestly that we don't currently have it, and ask about their target market and specific requirements. Do not imply that we can obtain it.
Original enquiry: [PASTE THE ENQUIRY]Be especially strict with this type. To make a reply sound positive, the AI may simply go along with the customer and write "yes, we support that", and once a promise like that is in an email it's very hard to take back.
Example: a price enquiry from arrival to sending
Take a fictional factory making stainless steel insulated bottles as an example. A German customer sends an enquiry through the website form: they want the 500 ml model printed with their company logo, 2,000 units, and ask about price and lead time.
- Gather and sort: Zapier syncs the form into the enquiry spreadsheet. It's classed as "price", with source marked "website" and country marked "Germany".
- AI drafts: using the price template above, with the tiered pricing for this model and how logo printing is charged written into the pricing rules. The AI's draft quotes the tier that matches 2,000 units, explains the logo printing cost, and asks about the port of destination and how many colours the logo has. The delivery time is written as [TO BE CONFIRMED].
- A person checks: the salesperson checks three things: whether the AI applied the right price tier; whether the logo printing surcharge is included; and whether the email makes any promise outside the pricing rules. Once that's confirmed, they look at the current production schedule, decide whether to give an approximate lead time, and replace [TO BE CONFIRMED].
Here the AI saves the time it takes to write a quotation email from scratch; whether the price is right and whether a lead time can be given is still the salesperson's call.
Which fields to include in your enquiry spreadsheet
When you gather enquiries into a spreadsheet in Step 1, settling the fields up front makes sorting, assigning and reviewing much easier later:
- Time received, source channel (website form / Alibaba.com / email)
- Customer country, company name
- Enquiry type (price / MOQ / certification / other)
- Product model, quantity, target price (leave blank if the customer didn't say)
- AI draft text or a link to it
- Assigned salesperson, review status, time sent
The "review status" column is key: the team can see at a glance which drafts nobody has looked at yet, so enquiries don't go cold sitting in the spreadsheet.
A checklist for salespeople reviewing AI drafts
- Prices, tiers and Incoterms match the current pricing rules
- No promises about delivery times, certifications or payment methods beyond what the rules allow
- The product model and specifications are the same ones the customer asked about
- The email asks for any key information the customer left out
- The reply is in the same language the customer wrote in, and the greeting and company name are correct
Common pitfalls with enquiry auto-replies
- Pricing rules left out of date: the AI only writes from the rules you give it. If the rules still have last quarter's prices, the drafts will quote the wrong price. When prices change, update the rules in your prompt at the same time.
- Feeding internal notes to the AI: internal information such as your floor price or margins could end up written into a reply. Only give the AI rules you're happy to share with customers.
- Using one template for every enquiry: if a certification enquiry gets the price template, the reply won't answer the question. Sort first, then pick the template.
- Looking only at how fast drafts are generated: a draft appears in seconds, but if reviews pile up with nobody handling them, your overall response time doesn't get any shorter. The review step needs a clear owner.
FAQ
Won't AI replies all sound the same? If you only give it one vague prompt, they can easily end up samey. We recommend writing a separate prompt template and set of real pricing rules for each of your common enquiry types (price, MOQ, certification). Drafts based on those templates will be much closer to how your business really works.
Should I tell customers an email was drafted by AI? What customers care about is whether the reply is prompt and the quote is accurate, not who wrote the first draft. Just make sure the content itself is accurate and professional.
If a customer writes in German or Spanish, can AI reply in their language? You can ask for this in the prompt: "reply in the language the customer wrote in". However, when the salesperson checks the draft, key information such as prices, specifications and Incoterms should be compared against the original, and if needed the draft can be translated back into Chinese for another read.
Should I use AI for very short enquiries that look like mass mailings? A general template is enough for these. The point is to politely ask about the product, quantity and target market, and only move into the formal quotation process once the customer has supplied more information. There's no need to spend time crafting a careful reply.
How should I write pricing rules so the AI doesn't get them wrong? Put them in a table or list them line by line: model, quantity range, price, Incoterms and validity each get their own column or line. Don't describe them in a paragraph. The more structured the rules, the less likely the AI is to mix up tiers when applying them, and the faster salespeople can check them.
To get started straight away, look at the details for each of these tools first: ChatGPT, Claude, Zapier. Beyond enquiries, common customer service questions on an independent store can be automated with the same "AI drafts, a person approves" approach; see how to auto-reply to store FAQs with AI. Once you have a lot of customer follow-up records, see Notion vs a spreadsheet for supplier and customer follow-ups to decide whether it's time to switch tools. To see how to choose every tool we've written about in one place, see how to build an AI tool stack for cross-border ecommerce.
Beyond enquiries and outreach emails, when your product information itself needs to be in several languages, see DeepL vs ChatGPT for product translation.
