A fashion retailer's Instagram and WhatsApp, answered around the clock
A women's fashion retailer sells almost entirely through Instagram DMs, comments and WhatsApp. A NUWIAX system now answers its customers day and night, from the live catalog and real order data, in each customer's own language. Every number on this page comes from this one client.
The problem
Is this in stock? What's the price? Where is my parcel? Customers ask the same questions hundreds of times a day, often late at night, and often by sending a screenshot or sharing a post instead of typing a product name. A keyword chatbot can't read a screenshot, doesn't know today's stock, and can't look up an order.
What we built
- Understanding: text, voice notes, photos and shared posts or reels, in any language the customer uses.
- Live lookups: price and stock from the store; order status by order number, behind a privacy gate so names, phones and addresses never reach the AI model.
- Product recognition: the design name and price printed on a screenshot are matched against every post the brand has published.
- Handover: frustration, unclear photos, rejected matches and unavailable items go to the team with a ticket.
- Operate: health checks every 30 minutes, a daily review of real conversations, alerts before a reply window closes, nightly backups and a live dashboard.
The hard parts
Shared posts. About one in five Instagram messages is a shared post or reel with a one-word question. Before we built for it, every one failed. The system now reads the post itself.
Screenshots, not studio photos. In a four-week manual review of 38 real customer photos across our retail systems, it named the right product 33 times, handed the other 5 to a person, and got none wrong.
Platform rules. Every send step is audited against Meta's messaging rules. When a platform restriction appeared, the system was adjusted within hours.