🌿 Vietnam E-Commerce Eco-Refund Engine
Traditional e-commerce refunds default to physical returns: the item ships back, gets inspected, restocked or discarded. For a premium Vietnamese platform, this created a compounding problem of logistics overhead, carbon emissions, and customer friction.
This Salesforce solution transforms the traditional, carbon-heavy refund process into a sustainable "Green" lifecycle — offering AI-generated Eco-Credit alternatives to reduce physical returns, logistics costs, and carbon emissions.
Features
🍏 Sustainable UX & Selective Routing
A Screen Flow with Intent Analysis logic identifies return scenarios like "Purchase by Mistake" and routes them through unique branching paths — promoting eco-friendly alternatives before ever triggering a physical return. This directly reduces return rates and transportation costs. Eco Credit is calculated based on the value of Customer Loyalty Level field on Refund Request records (Platinum, Gold, Silver and Standard).
🤖 AI-Generated Eco-Credit Offers
Einstein Prompt Builder uses a Flex Template to generate personalized, Vietnamese-language Eco-Credit offers for each refund scenario. The tone is professional and warm — designed for the cultural context of the platform's customer base.
🧑💻 Human Review & Automated Email
Before any offer reaches the customer, the AI-generated text surfaces on the agent's screen for review and audit. The agent saves only when satisfied — ensuring quality control while keeping the workflow fast and frictionless.
Upon agent approval, a customized email containing the AI-generated offer is sent directly to the customer. Simultaneously, the GeneratedOffer field on the Refund Request record is populated — creating a full audit trail for reporting.
Outcome
Reduced Physical Returns
Intent Analysis routing steers avoidable returns toward Eco-Credit alternatives — lowering reverse logistics costs and the platform's overall carbon footprint.
Faster Agent Resolution
A single button click initiates the entire AI offer workflow. Agents go from refund request to personalized customer email in minutes — not manual back-and-forth.
Localized, Culturally Relevant AI
AI offers are generated in Vietnamese with a tone calibrated for the platform's premium customer base — a detail that required deliberate prompt engineering to get right.
Reflection
This is one of my first attempts to utilize AI Prompt to solve a business solution. Engineering the Prompt Builder template for Vietnamese language and cultural tone was the hardest part. A generic prompt produces a generic offer. I had to iterate on the context structure: customer tier, order value, refund reason to get output that felt genuinely tailored rather than machine-generated. The human review step wasn't just a safety net; it was a deliberate trust mechanism for the agents using the tool.
