The Wednesday I opened WhatsApp to 127 unread vendor messages
I run a Shopify store in Pune. We sell handmade cotton bedsheets, block-printed duvet covers, and cushion covers—mostly to customers in Mumbai, Bangalore, and Delhi, but also through Flipkart and Meesho. I work with six suppliers: two in Jaipur, one in Surat, two local weavers, and a printing unit in Sanganer. During normal months, I get maybe twenty vendor messages a week. Order confirmations, delivery updates, the occasional pricing question.
Diwali prep is not a normal month.
This year, the first week of October, I opened WhatsApp Business to 127 unread messages. Flipkart Seller Hub had flagged eight late shipments. Meesho Supplier Panel was asking for updated GST invoices on four bulk orders. One Jaipur supplier wanted to renegotiate pricing on indigo-dyed fabric because cotton costs had jumped. Another supplier's courier partner had delayed three consignments, and I had customers messaging me directly asking where their orders were.
I spent two days typing replies. The same explanations, slightly reworded each time. "Please confirm the order quantity and delivery date." "Can you send the GST invoice for order #12845 by EOD?" "We're experiencing a delay due to courier issues; I'll update you by Friday." My thumbs hurt. I was still 40 messages behind.
That Thursday night, I decided to write five prompt templates—one for each type of message I was sending over and over. I tested them on the next 300 vendor messages across WhatsApp, Flipkart Seller Hub, and Meesho. I tracked which formula worked for which task, how much time each one saved, and where the AI replies sounded too stiff or missed the tone I needed.
Template one: order confirmations (the one that saved me an hour a day)
The most common message I send: confirming an order, restating the quantity, delivery date, and any special instructions. Before templates, I was typing this out twenty times a day, checking old WhatsApp threads to make sure I hadn't mixed up the block-print pattern or the thread count.
The prompt formula:
"You're confirming a wholesale fabric order with a supplier. Tone: polite, clear, businesslike but warm. Include: order number, item description, quantity, delivery date, payment terms. Keep it under 60 words. Avoid 'kindly' or 'please do the needful.'"
Then I'd paste the order details from my Shopify dashboard or Meesho panel.
What I used it for: Confirming orders with the Jaipur suppliers, the Surat weaver, and the Sanganer printing unit. Mostly on WhatsApp Business, sometimes copied into Flipkart Seller Hub messages when a supplier asked for written confirmation.
What worked: The replies were fast and specific. Claude Sonnet 4.5 kept the details in the right order and never added unnecessary pleasantries. I could send ten confirmations in fifteen minutes instead of an hour.
What didn't: Early attempts added "Looking forward to working with you" at the end of every message, even to suppliers I'd been working with for three years. I added "You've worked with this supplier before; skip introductory phrases" to the prompt, and it fixed itself.
I used this template 87 times across October. Saved me at least an hour a day during peak Diwali prep.
Template two: GST invoice requests (the one that needed the most editing)
Flipkart and Meesho both require GST invoices uploaded within 24 hours of dispatch. My suppliers know this, but during high-volume weeks, invoices get delayed. I was chasing four or five suppliers a day, trying to sound firm without being rude.
The prompt formula:
"You're asking a supplier to send a GST invoice for a completed order. Tone: polite but direct. Include: order number, dispatch date, invoice deadline. Mention that it's needed for marketplace compliance. Keep it under 50 words. No 'at the earliest' or 'kindly.'"
What I used it for: Meesho Supplier Panel messages and WhatsApp follow-ups when invoices were overdue. Occasionally adapted for Flipkart Seller Hub if a supplier hadn't uploaded the invoice themselves.
What worked: The tone was better than my own rushed messages. I tend to sound annoyed when I'm chasing paperwork; the AI version stayed neutral.
What didn't: It kept defaulting to "Please send the GST invoice at your earliest convenience," which is exactly the phrase I'd banned. I had to edit about half the replies to say "by 5 PM tomorrow" or "by EOD today" instead. After I added "specify an exact deadline, not 'at your earliest,'" the formula improved.
Used this 52 times. Still edited about 30% of them, but faster than writing from scratch.
Template three: delivery delays (the one that sounded too corporate at first)
Courier delays are common during Diwali. Delhivery and Ecom Express get overwhelmed, and consignments sit in transit hubs for three or four extra days. I was explaining the same delay to customers and suppliers, trying to sound apologetic without overpromising a new date.
The prompt formula:
"You're informing a customer about a delivery delay caused by courier issues. Tone: apologetic, transparent, reassuring. Include: original delivery date, reason for delay, new expected date. Offer to provide tracking updates. Keep it under 70 words. Avoid corporate language like 'inconvenience' or 'regret.'"
What I used it for: WhatsApp messages to customers, and occasionally to suppliers when I needed to explain why I couldn't accept a new batch until the delayed one arrived.
What worked: Once I got the tone right, these replies felt human. The AI avoided the "We regret the inconvenience caused" phrasing that makes you sound like a telecom company.
What didn't: The first ten attempts were too formal. "We are currently experiencing delays due to logistical challenges" instead of "The courier's running late because of Diwali volume." I added "Write like you're messaging a regular customer, not issuing a press release," and it loosened up.
Used this 48 times. Edited about 20% to add specific tracking numbers or a WhatsApp update offer.
Template four: bulk pricing negotiations (the one that needed my voice)
One Jaipur supplier wanted to raise the per-metre price on indigo-dyed cotton from ₹340 to ₹385 because raw cotton costs had jumped in September. I understood the reasoning, but I also had Flipkart orders locked in at the old price. I needed to negotiate a compromise without damaging the relationship.
The prompt formula:
"You're negotiating a price increase with a long-term fabric supplier. Tone: respectful, collaborative, firm on budget limits. Acknowledge their cost pressures, explain your pricing constraints, propose a compromise (smaller increase, phased implementation, or bulk discount). Keep it under 100 words. Avoid 'unfortunately' or 'I'm afraid.'"
Then I'd add context: current price, proposed price, my budget ceiling, order volume.
What I used it for: WhatsApp messages to the Jaipur and Surat suppliers when they flagged cost increases. Also used a version for asking about bulk discounts when I was placing a large Diwali order.
What worked: The AI structured the negotiation well—acknowledged their position, stated mine, offered a middle path. It kept the tone warm without being a pushover.
What didn't: It couldn't read the relationship. I've worked with the Jaipur supplier since 2021; we've had tea together twice. The AI version was polite but distant. I edited every negotiation message to add a personal line—"Hope the monsoon didn't hit the workshop too hard" or "Let me know if the new loom's working out."
Used this 19 times. Edited every single one, but the structure saved me from starting with a blank text box.
Template five: return policy clarifications (the one I used least but needed most)
Meesho customers sometimes want to return an item because the block-print colour looks different on screen, or they ordered the wrong size. Meesho's return window is seven days; Flipkart's is ten. I was explaining the policy over and over, trying to sound helpful while also protecting my margins.
The prompt formula:
"You're explaining your return policy to a customer who wants to return a bedsheet. Tone: helpful, clear, not defensive. Include: return window, condition requirements (unused, original packaging), refund timeline. Mention that block-print colours may vary slightly from photos. Keep it under 80 words."
What I used it for: Meesho Supplier Panel messages and WhatsApp replies when customers asked about returns before ordering or after receiving the item.
What worked: The AI explained the policy without sounding like I was reading from a terms-and-conditions page. It acknowledged the colour-variation issue, which I'd been typing out manually every time.
What didn't: It was almost too accommodating. Early versions said "We're happy to accept returns for any reason," which isn't my policy. I tightened the prompt to specify "only if unused and in original packaging," and it corrected.
Used this 23 times. Barely edited—this was the template that worked best out of the box.
What I'd do differently next time
I didn't track time saved per message scientifically, but I can tell you I stopped working until 11 PM every night. The order-confirmation template alone cut my daily admin from two hours to about 45 minutes. The GST invoice requests still needed editing, but I wasn't starting from scratch.
The biggest lesson: the first version of every template was too formal. I had to add "Write like you're messaging a supplier you know, not a corporate email" to almost every prompt before the tone felt right. Once I did that, the replies sounded like me—or close enough that I only edited for specifics, not for voice.
I also learned that negotiation and relationship messages need a human touch. The pricing template gave me a scaffold, but I had to add the personal details that acknowledged our history. AI can't know that I've been working with the same Jaipur supplier for three years, or that the Surat weaver's daughter just started college.
If you're handling vendor coordination during Diwali or wedding season, write your five most common messages as templates. Test them on ten real messages, edit the tone until it sounds like you, then use them for the next 200. You'll still edit some, but you won't be retyping the same order confirmation at 10 PM on a Wednesday.
I used Kryotta's workspace to store all five templates in one project, so I could switch between Claude Sonnet 4.5 for the fast confirmations and Gemini Flash when I needed a slightly warmer tone for customer-facing delays. The compare mode helped me see which model handled the negotiation scaffold better (Claude won). If you're juggling Flipkart, Meesho, and WhatsApp vendor messages, it's worth setting up a dedicated workspace so you're not hunting for the right prompt in your chat history at midnight.
Questions people ask
Do the AI replies sound robotic to suppliers?
The first drafts did. Once I added "Write like you're messaging someone you know" and banned phrases like "at your earliest convenience," they sounded human enough that no supplier ever asked if I was using AI. I still edited about 20–30% of messages to add specific details.
Which template saved the most time?
Order confirmations. I was sending 15–20 a day during peak Diwali prep, and that template cut the time from about 90 minutes to 20 minutes. The GST invoice requests saved time too, but needed more editing.
Can I use these templates for Amazon.in vendor messages?
Yes. The formulas work across any platform—WhatsApp Business, Flipkart Seller Hub, Meesho Supplier Panel, Amazon.in Seller Central. You'll need to adjust the specifics (Amazon's return window is different, invoice upload deadlines vary), but the structure holds.
What if my supplier relationship is more informal—can I still use templates?
Absolutely. Just adjust the tone prompt. I added "casual, friendly, like you're texting a colleague" for one Sanganer supplier I've known since school, and the AI matched it. You'll still want to add personal details (family updates, local news), but the template handles the logistics.
If you're staring at 100+ vendor messages and wondering how you'll get through Diwali prep without hiring an assistant, try writing five templates for your most common replies. Test them, edit the tone, then let the AI handle the repetitive structure while you add the human details that matter. You can set up a workspace and test different models at Kryotta—it's faster than switching between three browser tabs at midnight.
