Results

While the project was used in a limited operational context, the impact on the specific workflows it targeted was clear.

Order List Generation

~3 hrs → 2 min

Went from manual work to automated generation

Stockouts Decreased

Data-Driven

Reorder decisions based on actual consumption, not memory

Ingredient Cost Visibility

Instant

No more digging through invoices

Onboarding Simplified

Chat-Based

New staff could ask the chatbot instead of learning spreadsheets

Restaurants have rich operational data in their POS systems that goes almost entirely unused. The gap isn't analytics — it's connecting data sources that already exist.

A tool doesn't need to be sophisticated to be useful. Automating the math between “what we sold” and “what we need to order” is a simple concept with outsized impact.

Layer Technology
Frontend Next.js 14, React 18, TypeScript, Tailwind CSS, Framer Motion
API Next.js API Routes
AI OpenAI GPT-4 (chat + intent routing)
OCR Google Vision API
Backend Google Apps Script (V8 runtime)
Database Google Sheets
POS Integration Givex (extensible parser pattern for Square, Toast)
Deployment Vercel (frontend), Google Cloud (GAS + Sheets)

The project validated the core idea: connecting POS sales data to inventory tracking eliminates the manual guesswork in restaurant ordering. A future version could evolve toward:

  • trending_up

    Demand Forecasting

    True demand forecasting using historical sales patterns and seasonality

  • apartment

    Multi-Location Support

    Centralized inventory views across multiple restaurant locations

  • send

    Direct Supplier Ordering

    API integrations to auto-submit orders directly to suppliers

  • smartphone

    Mobile-First Interface

    For managers doing stock checks on the floor

  • database

    Database Migration

    Migration to Supabase/PostgreSQL for better performance and multi-user concurrency