WhatsApp AI assistant for customer queries and lead capture

The problem

For many Indian businesses, WhatsApp is the front desk. Customers ask for prices, availability, delivery times and order status at all hours. A small team answers the same fifty questions repeatedly, misses messages at night, and loses enquiries that were never written down anywhere.

A generic chatbot with fixed menus frustrates people. A language model on its own may answer confidently and wrongly. The useful middle is an assistant that answers from your approved information, knows when to stop, and hands the conversation to a person with the context attached.

How it works

1. Receive. The assistant connects to your number through the official WhatsApp Business Platform, so messages, templates and opt-ins follow WhatsApp's rules.
2. Understand. A language model reads the message in English, Hindi, Marathi or mixed text and works out the intent: product query, price, order status, complaint, or something else.
3. Answer from your content. Using retrieval over your catalogue, price information and FAQs (see RAG), it answers and can share product details, images and links. If the answer is not in your content, it says so.
4. Capture the lead. It asks for a name, need and city, confirms the phone number it is talking to, and writes the lead to your CRM or Google Sheet with the conversation summary.
5. Look up orders. With the customer's order number, it reads status from your order system and replies with the current stage and expected delivery.
6. Hand over. When the customer asks for a person, is upset, or the topic is out of scope, a staff member is notified and sees the full chat history.

Inputs, AI action and output

Input	What the AI does	Output
Customer message in any supported language	Detects intent and language	Routed request
Catalogue, prices, FAQs	Retrieves relevant content and answers	Grounded reply with details
Enquiry details given in chat	Extracts and stores lead fields	Lead record in CRM
Order number	Looks up order status	Status reply
Escalation trigger	Notifies staff with context	Human takeover

Integrations

We build on Google Cloud with Gemini models and connect to the WhatsApp Business Platform, your CRM or Sheets, your order or billing system, and an inbox where staff can reply. Teams who want quotes from chat can pair it with the AI quotation agent. For the wider approach, see generative AI assistants.

Guardrails and human review

• Answers come only from the content you approve. Prices and policies are not improvised.
• We use the official platform, with opt-in handling and approved message templates for business-initiated messages. We do not use unofficial WhatsApp automation, which risks your number being blocked.
• Sensitive topics such as refunds or complaints can always route to a person.
• Phone numbers and chats are stored securely under your retention rules. Confirm DPDP Act obligations with your adviser.
• Unanswered questions are logged weekly so you can add the missing content.

Rollout plan

We first gather your catalogue, FAQs and a sample of real chats to see what customers ask. A prototype is tested on an internal number with your staff, using a list of real questions to measure quality. We then connect the CRM and order lookup, switch on live traffic for a limited set of topics, and expand as accuracy holds. After go-live, we review gaps and keep content current under an AMC.

Data you need to start

• Your catalogue, price information and delivery or service terms in a shareable form.
• An FAQ list, or 100 or more real customer chats to build one from.
• A WhatsApp Business account, or a plan to set one up on the official platform.
• Read access to the order system if you want status lookups.
• A named person who owns the content and handles escalations.

A hypothetical walkthrough

This is an illustration, not a real customer. At 9:40 pm someone writes, in Hindi mixed with English, asking whether a particular steel cabinet is available in grey and what delivery to Nashik costs. The assistant finds the product, replies with the colour options, the price and the delivery terms from your content, and sends a photo. The customer asks to talk to someone about a bulk order. The assistant records the name, quantity and city, tells the customer a team member will reply in the morning, and creates a lead. At 9:30 am your salesperson sees the lead and the full chat.

What changes for your team day to day

Repetitive questions stop interrupting the team, and nothing arriving at night is lost. Staff handle fewer but more valuable conversations, and each starts with the history already visible. Someone needs to spend a little time each week reviewing the questions the assistant could not answer and adding content, which steadily widens what it can handle.

Who it is for

Retailers, distributors, clinics, institutes and service businesses where a large share of enquiries arrive on WhatsApp and are repetitive. See our view on build vs buy AI for SMEs to judge whether custom is right, or contact us.