Choosing live chat for a Shopify store in India

What do shoppers actually ask in chat?

On most D2C stores the same handful of questions make up the bulk of chats: is COD available to my pin code, when will it arrive, how do returns work, which variant suits me, and where is my order. The first four are pre purchase and decide whether a sale happens. The last is support.

Separate the two. Pre purchase questions should be answered in seconds, at any hour. Order status questions belong with your order system and your team.

Should chat be AI or human?

Both, with a clear line between them. An AI agent is good at answering from your published content and catalogue at three in the morning. It should not take payments, process refunds, promise delivery dates it cannot check, or argue with an upset customer.

The test of a good setup is the handoff: when the question crosses the line, the visitor is told plainly that a person will reply, and the person gets the whole conversation rather than a blank thread.

• AI: product questions, sizing, ingredients, policies already written down
• Human: orders, payments, refunds, complaints, anything unusual
• Both: a visible note of when a person is available

How should chat fit with WhatsApp?

Many Indian shoppers would rather continue on WhatsApp than wait on a website. Offer it as a handoff once the shopper has asked something that needs a person, rather than replacing website chat with a WhatsApp button that leaves the site. The website chat captures the question while the shopper is still on the product page.

What makes chat convert rather than just answer?

Showing the product, not describing it. When a shopper asks for a sunscreen for oily skin, a card with the right product, price and button converts better than three sentences. The chat should pull those cards from your live catalogue so prices and stock are current.

Opening at the right moment matters too. A chat that pops open on every page view is ignored; one that offers help when a visitor has hesitated on a product page is not.

What should you check before installing one?

Whether it reads your real catalogue and published pages, whether it can be told your hours and departments, how it hands off, what it stores, and whether you can bring your own AI model key or must pay per AI reply. Then test it with your ten most common questions.

IntellQ's chat agent answers from your published content and an imported catalogue from Shopify, WooCommerce or a CSV of up to 2,000 products, shows up to three product cards, and hands orders, payments, refunds and complaints to a person. It runs on your own OpenAI, Anthropic or Gemini key, or on Apavin Intelligence.

How do you set chat up in the first week?

Start by writing down answers to your twenty most common questions on your own site, in a FAQ or on the product pages, because that is what any AI agent will answer from. Set your real opening hours so the widget tells shoppers honestly when a person will reply. Decide which questions always go to a person, and who that person is during and after hours.

Then read every conversation for the first two weeks. You will find questions you did not expect and answers the agent phrased badly, and both are fixed by improving the content it reads rather than by tuning the model.

After that, check weekly: the share of chats handed to a person, how long those waited, and the questions the agent could not answer. That last list is your next set of FAQ entries.