AI Roadmap for Kovai Threads (fictional)

This report was generated automatically by software using AI. No human reviewed it before delivery. It is general information only, not professional, legal, tax or financial advice. Verify prices, vendor terms and legal requirements before acting.

Kovai Threads (fictional) · 2026-10-01 · AI Marga Roadmap Pro

Executive summary

Kovai Threads sells cotton kidswear from Tiruppur through its own Shopify store and on Meesho. The base is already in place: Shopify, the WhatsApp Business app and Zoho Books are running, and the team already uses some AI tools. The problems you named are ones that today's software handles well: return-to-origin (RTO) losses on cash-on-delivery (COD) orders, slow listing creation for new collections, the same customer questions every day, and stock decisions made by guesswork.

The biggest opportunity is COD confirmation. Every COD order on your Shopify store gets an automatic WhatsApp message, in Tamil or English, asking the buyer to confirm, cancel or switch to prepaid, and unconfirmed orders are held before packing. This is the most direct route to your goal of cutting RTO by a third. How close you get depends on how much of your RTO comes from fake or impulse orders rather than wrong addresses or delivery failures, so measure it from week one.

Second is faster collection launches. Sidekick, the AI assistant built into Shopify, is already included in your plan and can draft titles and descriptions; Photoroom can clean up phone photos of a whole collection in one batch. A person still checks sizes, fabric and label details, but starting from a draft instead of a blank page should cut listing time substantially (estimate). Third, the free Business Agent inside your WhatsApp Business app can answer routine questions about sizes, fabric, delivery times and returns, in Tamil and English.

What to do first: within the first month, choose one WhatsApp API platform (Interakt is the main suggestion; AiSensy is a lower-cost alternative), get COD confirmation live, and switch on the Business Agent in your existing WhatsApp Business app. Then move on to listings, Tamil versions of your key messages, and reorder planning from your sales history.

What to avoid for now: do not buy a full D2C suite or several WhatsApp tools at once, do not pay for an AI support agent until you see how many order-status questions remain after the free assistant is live, and do not paste customer names, phone numbers or addresses into free AI chat accounts. Amendments to the Consumer Protection (E-Commerce) Rules take effect on 1 January 2027 and cover how your own store shows discounts, handles complaints and displays product details. Build these into your listing work now and confirm what applies to you with a qualified professional.

Where your business stands today

This assessment is based only on what you entered in the intake form, so treat anything marked "not mentioned" as a question to check rather than a gap you definitely have.

Strengths

• Cloud tools across sales, messaging and accounts, so nothing needs replacing before you start.
• A clear, measurable goal (cut RTO by a third) that one tool change can be tested against.
• A team already comfortable trying AI tools.

Gaps to close first

• No automated COD check before dispatch.
• No combined, item-wise view of sales and stock across Shopify and Meesho.
• No written rules on what staff may paste into AI tools.
• Customer communication in Tamil depends on whoever is on the phone that day.

Your top AI opportunities, ranked

These seven are in the order to do them. The first three are low-effort and aimed straight at your goals; the last one is conditional.

1. Confirm cash-on-delivery orders automatically

Why it fits: High RTO on COD orders is your biggest named leak, and you want to cut it by a third. A confirmation step before dispatch stops many fake and impulse orders from ever leaving the building.

Day to day: When a COD order comes in on Shopify, the buyer gets a WhatsApp message in Tamil or English with the order summary and three buttons: confirm, cancel, or pay now and switch to prepaid. Confirmed orders go to packing. Unanswered orders get one reminder and are then held under a rule you set. Packing staff pack only confirmed COD orders.

Tools: Interakt or AiSensy (WhatsApp API platforms; choose one). fastrr by Shiprocket (Engage 360) adds AI voice calls for buyers who do not reply, worth considering later.

Effort: Low, about a week to set up.

Expected benefit (estimate): Fewer parcels shipped to buyers who never meant to accept them, fewer returned parcels to check back into stock, and some COD orders converted to prepaid. Confirmation will not fix RTO caused by wrong addresses or courier failures, so track reasons separately.

First three steps:

1. Export the last three months of Shopify COD orders, mark which came back as RTO and why (refused, unreachable, wrong address). This is your baseline.
2. Sign up with one platform on a new phone number (not the one on your WhatsApp Business app), connect Shopify, and get a COD confirmation template approved in Tamil and English. Ask the vendor to show the COD trigger working on a test order.
3. Add a clear WhatsApp opt-in line at checkout and agree the hold rule with the dispatch team.

Pitfall: Holding orders too long annoys genuine buyers. Keep the waiting time short and the message polite.

2. Create product listings and photos with AI

Why it fits: Slow listing creation is delaying every collection launch.

Day to day: Staff shoot each style on a phone in good light. Photoroom removes backgrounds in a batch and creates clean white-background images for Meesho and lifestyle images for your store. Sidekick, Shopify's built-in AI assistant, drafts the title, description and SEO text from your spec sheet. A person checks each draft against the physical label and size chart, then the same copy is adapted for Meesho. Canva handles collection banners.

Tools: Shopify (Sidekick), Photoroom, Canva.

Effort: Low, about a week.

Expected benefit (estimate): Writing time per product falls sharply when staff edit a draft instead of starting from blank, so each new collection can reach Shopify and Meesho sooner.

First three steps:

1. Create a spec spreadsheet with one row per style: fabric, sizes and age range, colours, care instructions, country of origin and the declarations printed on the label.
2. Try Photoroom's free tier on one existing collection and compare the results with your current photos.
3. Ask Sidekick to draft five listings, compare them with your best current listing, and save a standard instruction for tone and structure.

Pitfall: AI may invent details (a wrong fabric blend, "organic" or "chemical-free" claims) or shift a garment's colour in edited photos. Every listing must match the label.

3. Answer repeat questions with the WhatsApp Business app

Why it fits: You already use the WhatsApp Business app, and the same questions come in daily.

Day to day: The built-in Business Agent answers size, fabric, washing, delivery-time, COD and return-policy questions, day and night, from the information you give it. Staff handle complaints, refunds and order-specific questions. It answers from your business information, so it cannot look up a live order's status.

Tools: WhatsApp Business app (Business Agent), free up to a monthly allowance.

Effort: Low, about a week.

Expected benefit (estimate): Instant first replies at night and on Sundays, and fewer repeat questions reaching staff.

First three steps:

1. Write an FAQ document in English and Tamil: size by age, fabric, care, delivery times, COD, returns and exchanges.
2. Update the business profile, catalogue and quick replies.
3. Turn on the Business Agent, test it with real past questions in both languages, and check its answers daily for the first two weeks.

Pitfall: An out-of-date FAQ gives wrong answers. Update it the day a policy or price changes.

4. Communicate in Tamil as well as English

Why it fits: Your customers write in both languages.

Day to day: ChatGPT drafts Tamil versions of templates, FAQ answers, size charts and care instructions; a Tamil-fluent team member corrects them. The free Bhashini app turns a Tamil voice note into text. Canva's Translate feature converts size charts and creatives into Tamil.

Tools: ChatGPT, Bhashini, Canva.

Effort: Low.

Expected benefit (estimate): More queries handled in the customer's own language and fewer misunderstandings about sizes and returns.

First three steps:

1. Make a glossary of words that stay as they are (brand name, product names, size labels).
2. Draft Tamil versions of your ten most-sent messages and have them corrected.
3. Translate the size chart in Canva and check every number.

Pitfall: Machine Tamil can sound stiff or get meaning wrong. Nothing goes to customers unchecked.

5. Plan production and reorders from sales history

Why it fits: Every kidswear design multiplies into sizes and colours, so guesswork leads to stock-outs of fast sizes and piles of slow ones.

Day to day: Item-wise sales and stock from Shopify, Meesho and Zoho Books feed a Zoho Analytics dashboard showing fast and slow movers by style, size and season. The person who plans production reviews it weekly and sets reorder levels. ChatGPT can help read an export with customer columns removed.

Tools: Zoho Analytics (start on the free plan), ChatGPT.

Effort: Medium, about four weeks. It needs around twelve months of item-wise sales and recorded lead times.

Expected benefit (estimate): Fewer stock-outs on best-selling sizes and less dead stock, measured monthly.

First three steps:

1. Agree one SKU code format (style-colour-size) across Shopify, Meesho and Zoho Books.
2. Export twelve months of item-wise sales into one sheet.
3. Build a first fast/slow mover view and set reorder levels for your top styles.

Pitfall: Inconsistent SKU codes make the numbers useless. Fix the codes first.

6. Draft a month of social posts at once

Why it fits: Each collection needs Instagram and WhatsApp posts, and writing them one by one is slow.

Day to day: Once a month, the marketing person gives ChatGPT the collection list, offers and common customer questions, and gets a draft calendar with captions in English and Tamil. Canva's brand kit turns these into creatives; Meta AI helps with quick variations inside Instagram and WhatsApp. A person approves every post.

Tools: Canva, ChatGPT, Meta AI.

Effort: Low.

Expected benefit (estimate): Less time per post and more posts published on schedule.

First three steps:

1. Set up a Canva brand kit with logo, colours and a few example posts.
2. Draft next month's calendar around your next launch.
3. Fix a weekly approval slot.

Pitfall: Photos of children need written consent from a parent. Keep ads aimed at parents.

7. Order-status and returns agent (only if needed)

Why it fits: "Where is my order" and return questions need live order data, which the free assistant does not have.

Day to day: An AI agent on your WhatsApp API number reads order status from Shopify, answers delivery and return questions, and passes complaints to a person with the order details attached.

Tools: Interakt (AI Agents add-on), Shopify.

Effort: Medium, about three weeks.

Expected benefit (estimate): A share of order-status chats resolved without staff.

First three steps:

1. For two weeks, label order-status and return chats in the WhatsApp Business app and count them.
2. Write down your return, refund and exchange policy.
3. If volumes justify it, trial the agent with escalation to a person.

Pitfall: Refund and return decisions should stay with a person.

Department-by-department plan

Customer support

Current pain: The same questions arrive again and again on your WhatsApp Business number, in Tamil and English: which size for which age, whether the cotton shrinks, how long delivery takes, whether COD is available, how returns work. Each one is answered by hand, and order-status and return questions take longer because staff have to look the order up.

AI opportunity: Build support in layers and only pay for the next layer when the one before it is working.

1. The Business Agent in your WhatsApp Business app answers routine questions from your FAQ, at any hour.
2. Tamil versions of every saved reply and template, drafted with ChatGPT and corrected by a Tamil-fluent team member, so Tamil-speaking parents get the same quality of answer.
3. Only if order-status questions remain heavy: an AI agent on your WhatsApp API platform that reads live Shopify order data.

Complaint handling also needs a clear routine. From 1 January 2027, the amended Consumer Protection (E-Commerce) Rules require complaints to be acknowledged within 48 hours and resolved within a month. Use WhatsApp labels such as "Complaint – new", "Waiting for customer" and "Resolved" so nothing slips. Confirm the exact requirements with a qualified professional.

Tools: WhatsApp Business app (Business Agent), ChatGPT, Bhashini; later, Interakt's AI Agents add-on.

First step: For one week, label every incoming chat by type (size, fabric, delivery, COD, order status, return, complaint, other). The counts tell you what the FAQ must cover first and whether layer 3 is needed at all.

Metrics to track:

• Median first-response time on WhatsApp.
• Share of questions answered without staff involvement.
• Complaints acknowledged within 48 hours.

Marketing and content

Current pain: Creating listings and marketing content takes too long, which slows every collection launch on Shopify and Meesho.

AI opportunity: Turn listing creation into a repeatable pipeline: phone photos, then Photoroom for clean images, then a spec sheet, then Sidekick drafts in Shopify, then a person's check against the label, then Meesho. Run social content as a monthly batch, with ChatGPT drafting captions in English and Tamil, Canva's brand kit producing creatives, and Meta AI for quick variations.

Build the rules into the pipeline, not into memory:

• Online listings should match the physical label; the Legal Metrology (Packaged Commodities) Rules set what the label must declare.
• Claims such as "organic" or "eco-friendly" need proof under the CCPA's misleading-advertisement and greenwashing guidelines.
• No false urgency such as fake countdowns or "only 2 left" unless it is true, under the CCPA's dark-pattern guidelines.
• From 1 January 2027, when you announce a discount on your own store, the amended e-commerce rules require showing the lowest price from the 30 days before it. Confirm the details with a qualified professional.

Tools: Shopify (Sidekick), Photoroom, Canva, ChatGPT, Meta AI.

First step: For your next collection, note how many working days pass from final sample to live listing on each channel. This is the baseline your goal of launching faster is measured against.

Metrics to track:

• Days from final sample to live listing.
• Number of products with complete listings (all sizes, photos, care and label details).
• Posts published versus plan, and time spent per post.

Operations and delivery

Current pain: High RTO on COD orders, and production and reorder decisions made by guesswork.

AI opportunity: Two changes. First, COD confirmation on WhatsApp in Tamil and English, with unconfirmed orders held before dispatch and an option to switch to prepaid. Second, a monthly and then weekly review of fast and slow movers by style, size and colour, built in Zoho Analytics from Shopify, Meesho and Zoho Books data, so production and reorders follow actual demand.

Keep the two order channels apart when measuring. The COD confirmation tools in this roadmap connect to your Shopify store, so Meesho RTO should be tracked separately and not expected to change from this step.

Tools: Interakt or AiSensy, Shopify, Zoho Books, Zoho Analytics, ChatGPT.

First step: Build the RTO baseline. Take the last three months of Shopify COD orders and record, for each RTO, the reason if known (refused, unreachable, wrong address) and the delivery state or pincode. If most RTOs are refusals, confirmation will help a lot; if most are address problems, add an address check to the confirmation message.

Metrics to track:

• COD RTO rate on Shopify orders, weekly.
• Share of COD orders converted to prepaid.
• Stock-out incidents on top styles and value of slow-moving stock, monthly.

Recommended tools and indicative costs

Indicative prices from vendor pages on the date shown. Prices change, confirm with the vendor before buying.

How to choose between these tools

Start with what you already pay for. Sidekick is included in every Shopify plan, Zoho Books includes Zia for questions about your books, and the Business Agent in the WhatsApp Business app is free up to a monthly allowance. Use these fully before adding anything.

WhatsApp API platform: pick one, not two. COD confirmation needs a WhatsApp API platform on a separate number.

• Interakt is the main suggestion: it covers templates, a shared inbox, catalogue and WhatsApp payments, and offers an AI Agents add-on, so the same platform can later handle order-status questions.
• AiSensy has a free plan (without broadcasts) and a lower entry price, with its own AI Agent add-on. Choose it if budget is tight.

Decide in a trial: whichever one shows a working Shopify COD confirmation flow with Tamil templates, at a support level your team is happy with, wins. Per-message WhatsApp charges come on top of the platform fee and grow with order volume, so check them against your budget band.

Specialist COD tools: not yet. fastrr by Shiprocket (Engage 360) adds AI voice calls and RTO risk flagging but has no public pricing. GoKwik adds checkout and RTO risk profiling but charges per COD order on top of a monthly fee. Consider either only if a month of WhatsApp confirmation has not moved RTO enough.

Photos and design: Photoroom is for batch product-photo clean-up; Canva is for banners, social posts, size charts and Tamil versions of designs. Both have free tiers. Try them on one collection before paying.

AI assistant: ChatGPT (Free or Go) is enough for product copy, captions and Tamil drafts, provided no customer data goes in. Its Go plan may show ads. If several staff start using it daily with business documents, look at a business-tier assistant whose data is not used for training.

Dashboards: start Zoho Analytics on its free plan, since it connects to Zoho Books. Upgrade only when you hit its row or user limits.

Do not buy yet: a second WhatsApp platform, a full D2C support suite, paid social-media generators, or any tool that needs a developer. Your medium tech comfort suits no-code tools that one person can own.

Your 30/60/90-day plan

Roles below are suggestions: in a team of your size one person may hold two of them. Before week 1 ends, the owner should name one AI rollout owner (for example the operations lead) who keeps this plan moving.

Days 1–30: Stop the COD leak and switch on free tools

Week 1

• Owner: name the AI rollout owner and share the AI usage policy in this report with the team.
• Operations lead: build the RTO baseline from the last three months of Shopify COD orders, with the reason for each RTO where known.
• Customer support executive: label every WhatsApp chat by type for the whole week and start the FAQ document in English.

Week 2

• Operations lead: take trials or demos of Interakt and AiSensy, using a test order to see the COD confirmation flow working with Shopify. Pick one. Arrange a new phone number for it.
• Customer support executive and a Tamil-fluent team member: finish the FAQ in Tamil. Update the WhatsApp Business app profile, catalogue and quick replies.

Week 3

• Operations lead: submit the COD confirmation templates in Tamil and English for approval, connect Shopify and run test orders. Add the WhatsApp opt-in line at checkout (owner approves the wording).
• Customer support executive: turn on the Business Agent and test it with real past questions in both languages.

Week 4

• Go live with COD confirmation and the hold rule. Dispatch packs only confirmed or prepaid COD orders.
• Customer support executive: check the Business Agent's answers daily and correct the FAQ where it went wrong.

How to measure: every Shopify COD order receives a confirmation; weekly RTO rate compared with the baseline; median WhatsApp first-response time.

Days 31–60: Faster listings and Tamil everywhere

Week 5

• Catalogue executive: create the spec spreadsheet (fabric, sizes and age range, colours, care, country of origin, label declarations). Try Photoroom's free tier on one existing collection.

Week 6

• Catalogue executive: write a standard Sidekick instruction for tone and structure, then draft the next collection's listings. A second person checks each one against the label before it goes live on Shopify and Meesho.

Week 7

• Customer support executive with the Tamil-fluent team member: build the glossary, translate the ten most-sent messages and the size chart (using Canva's Translate feature), and check every figure.

Week 8

• Marketing executive: set up the Canva brand kit and draft next month's content calendar with ChatGPT.
• Owner and operations lead: review one full month of COD results. Adjust the hold rule, add an address check if wrong addresses are a big RTO reason, and decide whether AI voice calls are worth exploring.

How to measure: working days from final sample to live listing compared with your last collection; products with complete listings; RTO trend against the target of a one-third cut.

Days 61–90: Plan stock from data and prepare for the new rules

Week 9

• Production or purchase lead with the accounts executive: agree one SKU code format (style-colour-size) across Shopify, Meesho and Zoho Books, and export twelve months of item-wise sales.

Week 10

• Accounts executive: set up Zoho Analytics on the free plan, connect Zoho Books and the sales sheet, and build a fast/slow mover view by style, size and colour.

Week 11

• Production or purchase lead: set reorder levels for the top styles and start a short weekly stock review using the dashboard.

Week 12

• Owner and customer support executive: use the chat-label counts to decide whether an order-status AI agent is worth adding.
• Owner: list what the Consumer Protection (E-Commerce) Rules amendments taking effect on 1 January 2027 mean for your own store: prior-price display on discounts, complaint acknowledgement and resolution times, country of origin and GSTIN display, and a yearly dark-pattern self-audit. Confirm the details with a qualified professional.

Week 13

• AI rollout owner: review all metrics with the owner, update the FAQ and the AI usage policy, and set the next quarter's priorities.

How to measure: stock-out incidents on top styles; slow-moving stock; complaints acknowledged within 48 hours; RTO rate against the goal.

Step-by-step SOPs for your first two use cases

SOP 1: Confirming COD orders on WhatsApp

Purpose: Confirm every Shopify COD order before it is packed, so fake and impulse orders are cancelled or switched to prepaid instead of coming back as RTO.

Roles

• Owner: approves the message wording, the checkout opt-in line and the hold rule.
• Operations lead: owns the flow on the WhatsApp API platform (Interakt or AiSensy) and the weekly report.
• Customer support executive: handles replies that need a person.
• Dispatch team: packs only confirmed or prepaid COD orders.

Inputs: Shopify COD orders; approved confirmation and reminder templates in Tamil and English; the written hold rule; the buyer's WhatsApp opt-in from checkout.

Steps

1. When a COD order is placed, the platform sends the confirmation template automatically. Unless your platform can pick the language per customer, send one message with Tamil first and English below.
2. Confirm: the order is marked as confirmed (in Shopify or the platform dashboard, depending on how the integration works) and joins the packing list.
3. Cancel: the customer support executive cancels the order in Shopify the same working day.
4. Pay online: the buyer gets a payment link; once paid, the order is treated as prepaid.
5. No reply: the platform sends one reminder after the waiting time in your hold rule.
6. Still no reply: the order is held. During working hours, the customer support executive may send one personal message from the shared inbox.
7. At the end of the hold period, apply the owner's rule: cancel, or keep holding a little longer for repeat customers.
8. Every morning the dispatch team prints or views the list of confirmed and prepaid COD orders and packs only those.
9. Every Monday the operations lead records COD orders placed, confirmed, cancelled, unanswered, switched to prepaid, and RTO by reason, and compares them with the baseline.

Quality checks

• Weekly, read ten sent messages and check the name, items and amount match the order.
• Check that no unconfirmed COD order was dispatched.
• Check that opt-outs are honoured and no buyer got more than one reminder.

When the automation goes wrong

• Wrong order details in messages: pause the flow, fix the field mapping with the vendor, and message the affected buyers personally.
• Buyer replies with a question instead of tapping a button: answer from the inbox in working hours; add the question to the FAQ if it repeats.
• A genuine buyer was cancelled by mistake: contact them, restore the order, and note the cause so the rule can be fixed.
• Template rejected by Meta: reword it as a plain order update with no promotional language and resubmit.

SOP 2: Listing a new collection with AI drafts

Purpose: Get each new collection live on Shopify and Meesho faster, with every listing accurate to the physical label.

Roles

• Catalogue executive: photos, spec sheet and AI drafts.
• Reviewer (a different person, for example the production lead): checks every listing against the garment and label.
• Owner: decides on any product claim the reviewer is unsure about.

Inputs: Final samples with labels; the spec spreadsheet; a phone; Photoroom; Shopify with Sidekick; your standard tone instruction; the size chart.

Steps

1. Fill one spec-sheet row per style from the physical label and size chart, never from memory.
2. Photograph each style in daylight on a plain surface: front, back and a fabric close-up, using the same angles every time.
3. Process the batch in Photoroom: white backgrounds for Meesho, lifestyle backgrounds for your store. Hold each image next to the garment and check the colour.
4. In Shopify, create the product, give Sidekick the spec row and your standard instruction, and ask for a title, description and SEO text.
5. Edit the draft: remove anything that is not in the spec row.
6. The reviewer runs the quality checks below and signs off.
7. Publish on Shopify, then adapt the title and description to Meesho's listing format using the same facts.
8. Record the publish date so you can track days from final sample to live listing.

Quality checks

• Fabric, sizes, age ranges and care instructions match the label and size chart exactly.
• Country of origin and the other label declarations are shown correctly.
• No unproven claims such as "organic", "eco-friendly" or "chemical-free", and no urgency lines that are not true.
• Photo colours match the garment.
• From 1 January 2027, any discount on your store shows the correct prior price as the amended e-commerce rules require.

When the AI output is wrong

• A wrong fact: correct it, then add the correct fact to the spec row or standard instruction so it does not repeat.
• The same mistake keeps coming back: shorten the instruction and add one example of a good, approved listing.
• An edited photo changes the garment: use background removal only for that style, with no generated background.
• Never publish a draft "to fix later". If the reviewer is unsure, the listing waits.

Ready-to-use prompts for your team

Copy a prompt, replace everything in [square brackets], and paste it into ChatGPT or another approved AI assistant. Never paste customer names, phone numbers, addresses or payment details. Remove them first; an order number on its own is enough to identify an order inside your own systems.

Customer support

Answer a parent's question in both languages (use when a question is not yet covered by the WhatsApp Business Agent):

You help Kovai Threads, a cotton kidswear brand, reply to parents on WhatsApp. Write a short, friendly answer to this question: [customer question, with name and number removed]. Use only these facts: [relevant lines from our FAQ]. If the facts do not answer the question, say a team member will reply soon. Give the answer in English, then in simple spoken Tamil. Keep each version under 60 words.

Acknowledge a complaint (use for wrong size, damaged item or late delivery complaints):

Write a polite WhatsApp reply acknowledging a complaint about [issue] on order [order number]. Apologise once, explain the next step: [what we will do and by when], and say when they will hear from us. Do not promise anything I have not written here. Give English and Tamil versions.

Understand a Tamil message quickly (use when the person on duty is not comfortable reading Tamil):

Translate this customer message from Tamil into English, then summarise in one line what the customer wants and how urgent it is: [customer message with name, number and address removed]

Marketing and content

Draft a product listing (use when Sidekick's draft needs a second option):

Write a Shopify product listing for a kids' cotton garment. Use only these facts and add nothing else: Product: [name]. Fabric: [exact fabric from label]. Sizes and age range: [sizes]. Colours: [colours]. Care: [care instructions]. Country of origin: [country]. Write a title under [number] characters, three short benefit bullets, a 60-word description in warm, simple English for parents, and a meta description under 155 characters. Do not use "organic", "eco-friendly", "chemical-free", "best" or any urgency line.

Adapt a listing for a marketplace (use after the Shopify listing is approved):

Rewrite this approved listing for a marketplace product page: a shorter title, a plain list of features and no brand story. Keep every fact exactly as written and add no new ones. Listing: [approved Shopify listing]

Plan a month of posts (use once a month, before a launch or festival):

Plan [number] Instagram posts and [number] WhatsApp Status updates for [month] for Kovai Threads, a cotton kidswear brand from Tiruppur. Products to feature: [collection names and key facts]. Offers: [offers with exact dates, or "none"]. Parent questions to answer in posts: [three to five questions]. For each post give the date, format (photo, carousel or reel idea), an English caption, the same caption in simple Tamil, and the photo needed. No urgency lines and no claims beyond the facts given.

Operations and delivery

Write the COD confirmation message (use when setting up or improving the template):

Write a WhatsApp order-update message asking a customer to confirm a cash-on-delivery order. Use the placeholders [name], [order number], [items] and [amount], which our WhatsApp platform fills in. Ask them to tap Confirm, Cancel or Pay online now. Under 50 words, polite, no promotion. Give an English version and a simple Tamil version.

Find fast and slow sizes (use monthly with an export that has no customer columns):

I am attaching item-wise sales for a kidswear brand for the last [number] months. Columns: SKU, style, size, colour, channel, month, units sold, units in stock. There is no customer data. List the ten fastest and ten slowest style-size combinations, the sizes that sell out first in each style, and any monthly pattern. Suggest reorder levels for the top ten styles, assuming a lead time of [number] weeks. Show a table and flag anything that looks like a data error.

Learn from return reasons (use monthly):

Here are return and exchange reasons from [period], with customer details removed: [list of reasons]. Group them into themes such as too small, too big, colour different from photo, quality and other. Count each theme and suggest specific changes to our size chart, photos or descriptions.

Data protection and risks (DPDP awareness)

Kovai Threads handles customer personal data every day: parents' names, phone numbers, delivery addresses, order histories, COD and payment status, and WhatsApp chats, plus any photos of children that parents send you. Under the Digital Personal Data Protection Act, 2023 you are a Data Fiduciary for this data. Shopify, your WhatsApp platform and any AI tool that receives customer data act as your Data Processors, and you remain responsible for what they do with it. There is no turnover threshold. The main obligations under the DPDP Rules, 2025 (notice, consent, security, breach intimation, children's data and people's rights) apply from 13 May 2027; watch for any amendment that moves this date.

What must not go into AI tools

• Customer names, phone numbers, addresses or payment details, including order exports with customer columns.
• Whole WhatsApp chat exports.
• Photos of children sent by parents.

Remove these before pasting anything into ChatGPT or another assistant. Product details, anonymised sales figures and questions with personal details removed are fine.

Children's data

Your buyers are parents, but some data can still be about children: a child's name for a birthday offer, a photo for a review, an age for sizing. Children's data needs verifiable consent from a parent or guardian, and tracking, behavioural monitoring or targeted advertising directed at children is prohibited. Keep your marketing aimed at parents, and get a parent's written consent before reposting any photo of a child.

Consent and messaging

• Get a clear WhatsApp opt-in at checkout before sending COD confirmations or cart messages. Pre-ticked boxes do not count.
• Order updates and marketing are different purposes. Do not start sending promotions to numbers collected only for order updates without fresh consent.
• Honour opt-outs quickly. Promotional SMS or calls also fall under TRAI's commercial communication rules.

Checks before choosing a vendor

• Is your data used to train their AI models? Business tiers usually say no.
• Where is the data stored, and are security terms written into the contract?
• Can you set how long chats and uploads are kept, and delete them?
• Will they delete your data when you leave?

Where a person must review AI output

• Refund, return and cancellation decisions.
• Replies to complaints.
• Every listing before it is published, for label accuracy and claims.
• Every Tamil translation before it reaches customers.

Simple habits to start now

Keep a one-page register of which tool holds which customer data and who can access it. Write a one-page breach plan naming who decides, who tells customers and who informs the Data Protection Board (detailed report within 72 hours) and CERT-In (certain cyber incidents within 6 hours).

This is general awareness information, not legal advice. Confirm your obligations with a qualified lawyer or company secretary.

AI usage policy template for your team

Copy this into a document, fill in the [square brackets], and have every team member read and sign it.

Kovai Threads – AI usage policy

1. Purpose We use AI tools to answer routine customer questions faster, confirm COD orders, draft product listings and social posts, translate between English and Tamil, and plan stock. AI drafts and assists; people decide. This policy sets out which tools we use, what data may go into them, and when a person must check the output.

2. Approved tools

Any other AI tool, including free apps and browser extensions, needs written approval from the policy owner before use with business information.

3. Data rules

• Never paste customer names, phone numbers, addresses, payment details or WhatsApp chat exports into ChatGPT, Meta AI or any assistant not listed above for that purpose.
• Never upload photos of children sent by customers into any AI tool.
• Allowed: product details, label information, size charts, anonymised sales figures and customer questions with personal details removed.
• Use company accounts only, never personal accounts, for business work.
• Do not download customer lists to personal phones or laptops.

4. Human review rules A named person checks before anything goes out:

• Every product listing, against the physical label and size chart.
• Every new or changed customer message template, in both languages.
• Every social post, for claims and urgency lines.
• Every complaint reply, refund, return or cancellation decision. AI may draft; a person decides.

5. Telling customers

• Our WhatsApp profile and automated replies say that some answers come from an automated assistant, and explain how to reach a person.
• We never present AI-generated product photos as photos of the real garment unless they show it accurately.
• We do not use AI to write fake reviews, fake urgency or claims we cannot prove.

6. When AI gets it wrong

• Stop using the wrong output, correct it, and if it reached a customer, contact them with the correct information.
• Tell the policy owner the same day, with an example.
• The policy owner updates the FAQ, template or instruction so it does not happen again.
• If customer data may have gone to the wrong place, tell the policy owner immediately so the breach plan can be followed.

7. Policy owner [Name, role] owns this policy, approves new tools, keeps the list of tools and the data each one holds, and answers staff questions.

8. Review This policy is reviewed every six months, and whenever a new AI tool is added. Next review date: [date].

Signed: ____________ Date: ____________

Vendor shortlist and questions to ask

Indicative costs for each tool are in the cost table. Keep Shopify and Zoho Books as they are; the choices below are for the tools you add.

WhatsApp API platform (COD confirmation, later order support)

Specialist COD and RTO tools (only if WhatsApp confirmation is not enough)

Product photos and design

AI assistant

Indian-language support

Dashboards

Questions to ask every vendor

1. Where is our data stored, and can you put that in writing?
2. Is our data or our customers' data used to train your AI models?
3. Will you sign a data-processing agreement with security terms, ready for the DPDP Act and Rules?
4. Can your templates and AI replies handle Tamil as well as English? Please show us.
5. Can you show the COD confirmation flow working with a test order from our Shopify store?
6. What are your support hours, and is support available on WhatsApp or phone?
7. What is the minimum contract length, and what happens if we cancel mid-term?
8. Can we export all our chats, contacts and templates if we leave, and will you delete our data afterwards?
9. Which charges come on top of the plan (per-message fees, AI add-ons, taxes), and how much notice do you give before price changes?
10. Who on your side can see our customer chats, and is that access logged?

Government schemes worth checking

Scheme details change often. Check the official page and your eligibility before applying.

Next steps this week

• ☐ Owner: name one AI rollout owner, share the AI usage policy from this report, and tell the team that customer names, numbers and addresses no longer go into AI chat tools.
• ☐ Operations lead: export the last three months of Shopify COD orders and mark each RTO with its reason (refused, unreachable, wrong address). This is the baseline for your goal of cutting RTO by a third.
• ☐ Customer support executive: create WhatsApp labels (size, fabric, delivery, COD, order status, return, complaint) and tag every chat this week.
• ☐ Customer support executive: write the first English FAQ draft covering size by age, fabric, care, delivery times, COD and returns, and ask a Tamil-fluent colleague to start the Tamil version.
• ☐ Operations lead: book trials or demos with Interakt and AiSensy and ask each to show COD confirmation on a Shopify test order.
• ☐ Owner: arrange a separate phone number for the WhatsApp API platform and draft the opt-in line for checkout.
• ☐ Catalogue executive: note how many working days your most recent collection took from final sample to live listing on Shopify and Meesho.

Important notes

This report was generated automatically by software using AI. No human reviewed it before delivery. It is general information only, not professional, legal, tax or financial advice. Verify prices, vendor terms and legal requirements before acting.