WhatsApp support agents for Indian customers: Hindi, Gujarati and Hinglish

What the inbox actually looks like

A direct to consumer snacks brand in Ahmedabad sells khakhra, chikki and namkeen through its own website, Amazon and a few modern trade stores. Most customer questions arrive on WhatsApp. On a normal day the support team of four reads messages like these.

• "mera order kab aayega, 3 din ho gaye"
• "maru parcel ma ek packet futelu hatu, refund joie che"
• "Do you ship to Dubai?"
• "methi khakhra ma gluten che ke nahi"
• A photo of a damaged box with no text at all

Half are in Hindi or Gujarati typed in English letters. Some switch language mid sentence. Many refer to an order without giving the order number. During Diwali, the volume triples and replies slip from minutes to a day.

A WhatsApp support agent can answer the routine questions straight away, look up orders, collect what a person needs for the harder cases, and pass those on.

Decide what the agent handles

Start by reading two weeks of real conversations and sorting them. In most D2C inboxes, the groups look something like this.

• Order status: where is my order, when will it arrive. The agent can answer this fully by looking up the order.
• Product questions: ingredients, shelf life, allergens, sizes. The agent can answer from your product sheets.
• Shipping and policy: delivery areas, cash on delivery, return windows. The agent answers from your policy pages.
• Problems: damaged items, missing items, wrong orders, refunds. The agent collects details and photos, then hands over.
• Everything else: complaints, bulk orders, partnership requests. Hand over straight away.

Write this list down. Set the agent to answer the first three groups and to hand over the last two.

Handle languages the way customers write

Current language models understand Hindi and Gujarati reasonably well, including transliterated text typed in Latin letters and mixed sentences. The harder part is replying in a way that feels natural.

Reply in the customer's style

A simple rule works well: reply in the language and script the customer used. If they write "mera order kab aayega", answer in Hinglish in Latin letters, and keep formal Hindi in Devanagari for customers who write that way. Test this with your own team, because what reads naturally in Surat may differ from Jaipur.

Keep a glossary

Product names, sizes and policy terms should stay consistent across languages. "Methi khakhra 200 g" should never become a translated phrase. Keep a short glossary in the agent's knowledge source with your product names and the words customers use for them.

Know the weak spots

Regional slang, heavy abbreviations and voice notes are harder. If you allow voice notes, the agent first converts them to text, and accuracy for Gujarati speech varies with background noise and accent. Ask the customer to confirm the key details when the agent is unsure.

Connect it to orders

An agent that can only say "please share your order number" is not much help. Connect it to the systems that hold order data: your store platform such as Shopify, WooCommerce or ShopenUp, your shipping partner's tracking, and your helpdesk.

When a customer asks about an order, the agent should match them by the phone number they are messaging from, find recent orders, and confirm which one they mean. Then it gives the status and expected delivery date from the courier. It should never show order details to a number that does not match the order.

Follow WhatsApp's rules

The WhatsApp Business Platform has its own rules, set by Meta, and they change from time to time, so check the current version before you launch. As of 2026, the main points to plan around are:

• You can reply freely within a customer service window that opens when the customer messages you.
• Messages you start, such as delivery updates or a follow up after the window closes, must use templates that Meta approves in advance.
• Meta charges for some message types, and the pricing model has changed in recent years.
• Customers must have agreed to receive messages from you, and opting out must be easy.

Customer phone numbers and chat history are personal data under the Digital Personal Data Protection Act, 2023, so keep chat logs secure and only as long as you need them.

Design the hand over carefully

The hand over to a person needs as much design as the answers. Four details matter.

• Hand over early for problems. A customer with a damaged parcel wants a person. The agent should collect the order, the photo and a short description, then say clearly that a team member will reply and roughly when.
• Pass the full context. The support person should see the conversation, the order and what the agent already said, so the customer does not repeat themselves.
• Let the customer ask for a person at any point, in any language. "Kisi insaan se baat karni hai" should work as well as "talk to agent".
• Never promise what a person must decide. Refunds, replacements and goodwill gestures are for your team to approve. The agent can say a request has been raised, and only your team can say it is approved.

Limits to plan for

The agent will sometimes misread a mixed language message, especially short ones with abbreviations. It will occasionally reply in the wrong script. Review a sample of conversations each week in the first month and adjust the instructions and glossary.

It will also not fix slow courier partners or unclear return policies. If your policy says "returns at our discretion", the agent cannot make that clearer than it is.

Checklist

• Two weeks of real chats sorted into groups
• Clear list of what the agent answers and what it hands over
• Reply language rule and a product glossary
• Order lookup matched to the customer's phone number
• WhatsApp templates approved for messages you start
• Hand over that passes full context to your team
• Weekly review of a sample of conversations

Support agent shows how this works on AksharAI, including the WhatsApp channel and the hand over to your team.