How AI Automates WhatsApp Product Questions, Stock Checks, and Order Placement

Retail customers don't browse a WhatsApp catalog the way they browse a website. They ask a question: "Do you have this in blue?" "What sizes are left?" "Can I get it under $50?" A generic chatbot answers these badly, because it's either guessing from a static FAQ or hallucinating an answer that sounds plausible and isn't. For a retailer, a wrong stock answer isn't a UX nitpick — it's a cancelled order and an angry customer.

The core problem: retail conversations aren't scripted

A real product conversation branches constantly. A customer asks about one item, then pivots to a cheaper alternative, then asks if it comes in a different color, then decides to buy. Any system that treats this as a single-turn Q&A will break the moment the second message arrives. What's needed is a conversation engine that keeps track of what's been discussed, resolves ambiguous references ("the black one" — black what, from three messages ago?), and never answers a stock or price question from memory instead of the actual catalog.

What "grounded in the catalog" actually means

The distinction that matters is where the answer comes from. A system built for retail should route every product question through a real catalog lookup — budget filters, brand filters, fuzzy name matching — and return the current price and stock state, not a plausible-sounding one. When a customer asks for "something under $50," that's a structured query against real inventory, not a paragraph an LLM improvises. If a product genuinely doesn't exist in the catalog, the system should say so plainly instead of inventing a SKU that doesn't exist.

Variant handling is where most bolted-on chatbots fall apart. A shirt with five colors and four sizes isn't one product — it's up to twenty. When a customer says "the blue one" after being shown a list, the system needs to hold that pending context and resolve it on the next message, not restart the conversation from zero. Done right, this takes one clarifying question at most, not a wall of every possible combination.

From question to confirmed order, without extra steps

The order itself needs the same discipline. A customer who says "I'll take it" shouldn't have to fill out a form — the conversation should collect delivery address and payment method in the same thread, show a full order summary, and require an explicit "yes" before anything is created. That confirmation gate matters: without it, a system that's too eager to place orders creates as many support tickets as it prevents.

Two details separate a retail-grade assistant from a generic one:

• Discount resolution at the point of lookup. If a loyalty discount or promotion applies, it should appear the moment the product is shown — and stay consistent through address collection, payment, and final confirmation. A customer who sees one price at lookup and a different one at checkout stops trusting the channel.
• Returning-customer recognition. A repeat buyer shouldn't re-enter their address and payment method every time. When those details are already on file, checkout should collapse to a single confirmation message instead of four rounds of data collection.

Why this belongs on WhatsApp specifically

Retail buyers already research on WhatsApp — forwarding product links to family, asking a store's number a quick question before a purchase they've mostly already decided on. That's high-intent traffic. The businesses losing the most here aren't the ones without a WhatsApp presence — they're the ones whose WhatsApp number goes to a shared inbox that gets checked twice a day. An automated, catalog-grounded assistant that responds in under a minute converts a meaningfully higher share of that traffic, simply by being present when the intent is highest.

None of this requires the customer to know they're talking to AI, and it doesn't require the retailer to rebuild their catalog in a new system — it should sit on top of the product data that already exists, whether that's an e-commerce backend or a spreadsheet-driven inventory.

See how Sellatrix works for this industry →