How Xica Moda Mineira started to sell while the team sleeps and still wins back customers who vanished.

Context

Xica Moda Mineira is a women's fashion brand with a strong digital presence. It sells via WhatsApp, marketplaces, and a physical store. It runs campaigns, launches collections, opens clearance sales. And like every fashion brand, it lives in cycles: a launch explodes into messages, then drops off, and starts again.

Xica's sales operation was, like almost every brand that grew fast, heroic and exhausting. During campaigns, three phones active at once, more than 600 messages a day. Salespeople photographing a piece on their own body to send to the customer who asked. A customer coming in, asking the price, disappearing. A customer buying once, disappearing. A customer's birthday passing with no one remembering. Satisfaction surveys, repeat purchases, reactivation: all on the "someday" list.

And on top of that, a chronic technical problem: the ERP Xica used (C7) had no public API. That meant every integration had to be a hack: manual spreadsheet export, manual import somewhere else, double the rework.

The challenge

Xica didn't just want "a chatbot." It wanted the entire sales operation to stop depending on people being available. But selling clothes over WhatsApp has complexities a generic bot doesn't cover: the customer wants to see a photo of the piece, wants to know if her size is in stock, wants to compare it with another piece, wants to know the shipping cost to her city, wants to pay in installments.

And then there was the after-the-sale challenge, the part almost every fashion brand neglects. Xica was losing revenue not from a lack of new customers, but from failing to make the most of the customers it already had.

The solution

NexUnio built Xica's entire operation across three AI fronts, plus a system migration, plus a new product still in development.

24/7 Sales Agent. Takes over every new conversation that comes in on WhatsApp. Identifies what the customer wants, shows the product, calculates shipping to her ZIP code in real time, generates a payment link, logs the order directly in the new ERP. Closes the sale on its own, any hour, any day.

Re-engagement Agent. Monitors conversations that went cold. A customer who asked a question and stopped responding, a customer who added a product to the cart and disappeared: gets a personalized message at the right moment, with a coupon when it makes sense. Re-engages without being pushy.

Post-sale & Repurchase Agent. Automatic satisfaction survey after delivery. Birthday coupon on the right day. Scheduled reactivation for customers who've been inactive for 30, 60, 90 days, with a different message at each stage.

ERP migration. A complete migration pipeline from C7 (no API) to Bling (with a rich API). Catalog, multi-store inventory, POS with card readers: all migrated with zero operational downtime. From there, every integration became direct and stable.

NexTryOn (in development). A virtual fitting room powered by generative AI. The customer sends a photo and sees the outfit on her own body, eliminating the heroic step of salespeople taking a personal photo to send. An embeddable widget for the e-commerce platform.

The results

• Average ticket of R$ 451 held at scale: the agent sells with the same quality as an experienced human salesperson.
• Around 700 conversations/month tracked by the agent, with 9.4 messages per conversation on average.
• ERP migration with zero operational interruption: Xica moved from C7 to a system with native integration, and the AI took advantage of it immediately.
• Inactive customers get the right message at the right moment, with no one having to remember, and no annoying mass campaigns.

What this unlocks

Xica is the case where all three freedoms show up together:

Schedule: sells and serves customers at any hour of the day, any day of the week.

Scale: a launch campaign that used to turn into an all-nighter becomes a process absorbed by the system.

Memory: the knowledge of "how to serve a Xica customer" no longer depends on which salesperson picked up the conversation. The tone, the rules, the exchange policy, the way of treating people: all of it lives in the system.

And the virtual fitting room in development eliminates the last heroic step: no salesperson has to double as an improvised model anymore.