Client type
A clothing business that sells through WhatsApp. Most customer conversations start with the same questions: is this available in my size, what colours do you have, what does it cost, can you deliver. The client asked us not to publish their name.
The challenge
All of this ran through one or two phones. As message volume grew:
- Replies slowed down, especially in the evening, and customers moved on.
- Staff answered the same questions repeatedly, checking stock by hand each time.
- Answers were sometimes wrong because stock had changed since someone last looked.
- Only the person holding the phone could see the conversation, so handovers were messy.
What we built
We built a shared sales inbox connected to the business’s WhatsApp number, with an AI assistant that answers from live data.
Shared team inbox. Every WhatsApp conversation appears in a web inbox the whole team can use. Conversations can be assigned, labelled and handed over, with full history visible.
Auto-replies grounded in live data. When a customer asks about a product, size, colour or price, the assistant looks up current inventory and sales data and replies with what is actually available. It does not guess. If the data doesn’t answer the question, it says so and hands over.
Human handover. Anything outside the assistant’s scope (complaints, custom requests, payment issues, or anything it is unsure about) goes to a team member, flagged in the inbox. Staff can take over any conversation at any time.
Guardrails. The assistant is limited to the business’s own catalogue and policies, uses approved tone and wording, and never invents discounts, stock or delivery promises.
Sales context. The inbox shows the team relevant order and product information next to each conversation, so they can reply faster when they step in.
Tech stack
WhatsApp Business Platform (Cloud API) with webhooks, a Python service handling message routing and the assistant, a large language model with retrieval over the live inventory and sales database, PostgreSQL, and a web-based shared inbox for the team.
Outcome
Routine product questions are now answered immediately and accurately from current stock, at any hour. The team spends its time on conversations that need a person, and any team member can pick up a conversation with full context. Sales figures are confidential, so we describe the outcome qualitatively.
Could this work for you?
If your customers ask the same questions on WhatsApp or email and the answers live in your inventory, ERP or CRM, this pattern applies. See our AI workflow automation service, or read AI automation for finance teams for back-office examples.