Context
E-commerce operation managing 6 marketplaces simultaneously: Mercado Livre, Shopee, Magalu, Amazon, B2W and AliExpress. 7-figure monthly revenue. Ops team running across 3 different tools and 4 spreadsheets. Inventory drifting out of sync every single day. Customer service spent half its time re-entering order data between systems. Cancellations from overselling were constant.
The problem
Every marketplace has its own API, its own rules, its own way of reporting orders, inventory and tax data. Off-the-shelf SaaS (Bling, Tiny) covered part of it, but the client wanted custom dynamic pricing rules and automated copy generation per SKU. None of the available tools could do that. And on top of it: they wanted to own the code, not be locked into a third-party SaaS.
The call
Build from scratch. White-label platform, treated as a business asset, deployed on the client's own infrastructure. Modern stack to run lean and scale. AI applied inside the operational flow — not as an add-on layer.
The solution
We shipped it in 10 weeks:
- Integration layer with webhook + ETL per marketplace, covering orders, inventory, tax and logistics.
- Real-time inventory sync across the 6 channels, with optimistic locking to prevent overselling.
- Unified order panel with processing queue and per-channel routing rules.
- AI dynamic pricing layer based on competitor and inventory data (Claude API).
- Copy and image generation per SKU at scale, with a human-review pipeline before publishing.
The operation moved from 4 spreadsheets + 3 tools to a single platform. A single operator now manages what used to take 3 people.
The outcome
- 6 operations actively running on the platform.
- Overselling dropped from 4% to under 0.3%.
- SKU onboarding time dropped from 12min to 90s with AI.
- +R$ 36k/month in recovered operational efficiency (overselling + customer service time + faster onboarding).
- Client is on a continuous engagement 8 months in.
Why it worked
We didn't try to cover everything in the first release. We mapped the highest-pain areas (overselling + onboarding) and shipped those first. AI refinement came in weeks 6-8, after the base was stable. The kind of project that validated the AI-native thesis: strategy first, AI inside the flow where it generates ROI, code as the client's asset.
"I had given up on marketplaces. Kode delivered in 6 weeks what 2 agencies couldn't in 8 months."
— CEO of the operation