Parcel-Parse Ambiguous Retail Chats into Executable Logistic

• 

  Parcel, turn customer messages into orders.

• 

  IU Studio Demo Workspace - all features are fully functional

• 

  Orders Board gives the seller a simple pipeline view of every order: Needs Info, Awaiting Payment, Ready to Pack, and Shipped

• 

  From customer chat, turned into invoice

• 

  Analytics page shows the business impact

• 

  The Inventory page shows how PARCEL connects orders to stock

Inspiration

I built PARCEL because most commerce tools assume every seller has a clean checkout flow. But many small sellers do not start with a polished storefront. They run their business through messy WhatsApp, Instagram, TikTok or Messenger.

That means every order creates invisible manual work: reading messages, identifying products, calculating totals, checking stock, creating invoices, packing items, and replying to customers. I wanted to use Gen AI for that overlooked part of commerce: the space between a customer saying “I want this” and a seller actually fulfilling the order.

What it does

PARCEL turns messy customer chats into structured business workflows.

From one customer message, PARCEL can extract:

• Customer name
• Items and quantities
• Delivery location
• Delivery date
• Missing information
• Total price
• Confidence score

Then it creates the next operational actions:

• Invoice preview
• Inventory reservation
• Packing checklist
• Ready-to-send customer reply

PARCEL is not just another chatbot. It helps sellers stop doing operations by hand inside their inbox.

How I built it

I built PARCEL using MeDo through conversation-driven development. I used prompts to create and iterate on the full product experience: the demo workspace, inbox dashboard, AI extraction panel, order workflow, invoice preview, inventory impact, packing checklist, and reply composer.

The core product flow is:

Unstructured customer chat
→ extracted order schema
→ validation and missing-info detection
→ invoice, inventory, packing, and reply actions

The most impressive part MeDo helped generate was the end-to-end transformation from an informal customer message into a usable retail operations workflow.

Challenges I ran into

The hardest part was making the app feel like a real operations tool instead of a generic AI demo.

I had to think carefully about what a seller actually needs after reading a message. It is not enough to extract text. The app needs to answer: Can this become an order? Is anything missing? Is there enough inventory? What should the seller say next?

Another challenge was designing the demo so judges could feel the impact quickly. The app needed to show the before-and-after moment clearly: chaotic chat becomes invoice, packing list, stock update, and customer reply.

Accomplishments that I'm proud of

I’m proud that PARCEL feels like a real product, not just a chatbot demo. The app does not stop at extracting text from a message. It turns that message into a complete retail workflow: order details, invoice preview, inventory reservation, packing checklist, and customer reply.

I’m also proud of making the demo easy to understand quickly. A judge can open the app, enter the demo workspace, and immediately see the core transformation: one messy customer chat becomes a structured, actionable order.

The biggest accomplishment was using MeDo to build an end-to-end product experience through natural language iteration. I was able to go from idea to deployed application with a working dashboard, realistic sample data, and a clear business use case.

What I learned

I learned that the strongest AI products are not always the ones with the most complex interface. They are the ones that turn messy human behavior into useful structure.

For PARCEL, the important insight was that small sellers do not just need “AI chat.” They need AI that understands a business workflow and helps them move faster with fewer mistakes.

What's next

Next, I would expand PARCEL with:

• Real integrations for WhatsApp, Instagram, and TikTok messages
• Payment link generation
• Multi-language order extraction

The long-term vision is simple: help small sellers turn conversations into commerce without drowning in manual operations.

Built With

• appmedo
• async
• auth
• autoprefixer
• biome
• carousel
• css
• date-fns
• edge
• embla
• flash
• form
• framer
• functions
• gateway
• gemini
• helmet
• hook
• lucide
• motion
• postcss
• postgresql
• react
• recharts
• router
• shadcn/ui
• sonner
• sse
• storage
• supabase
• tailwind
• typescript
• vite
• zod