SellerPilot AI

What if your best-performing ad is also your biggest inventory risk? In our synthetic example, a TV protector campaign shows 7x ROAS, but stock lasts only 10 days while replenishment takes 30. Increasing advertising just because ROAS looks good could leave customers ordering products you cannot fulfill. That's why I built SellerPilot AI. SellerPilot is a free, open-source Arabic-English marketplace operations copilot for Amazon.sa and Noon.sa sellers. It connects advertising results to inventory, calculates ROAS, ACoS, estimated contribution and days of stock cover with auditable TypeScript functions, and ranks campaign risks based on business evidence—not just revenue. Open a recommendation to see its SKU, attributed sales, remaining inventory and supplier lead time. A what-if simulator models the implications of a hypothetical +15% spend change, clearly labeling its assumptions rather than presenting a sales forecast. The seller stays in control: inspect the evidence, stage a proposal and export an approval-marked CSV. SellerPilot never quietly changes campaigns. Data Lab imports daily Amazon Sponsored Products reports with separate cost and VAT inputs, plus guided column-by-column mapping for Noon exports. A data-trust monitor identifies stale or duplicated reports, missing stock and conflicting product costs, blocking unsafe approvals. Built during the hackathon in React, TypeScript, Vite, Hono and Zod, with responsive Arabic RTL support, 50 passing automated tests and Lighthouse accessibility of 100/100 on mobile and desktop. The updated exactly 180-second Full-HD film uses new ElevenLabs v4 narration throughout, with no music, sound effects, ambient audio, or silent introduction. Eighteen cinematic scenes show enlarged, accurately framed interface captures, repaired opening and closing graphics, and animated synthetic inventory-risk evidence. It ends with the seller making the decision, not an LLM chat screen. Public examples are synthetic. Live demo: https://sellerpilot-ai-phi.vercel.app