The Problem: Your Inventory Has a Story. Your Business Just Isn't Listening.
Imagine running a small retail business. One product is almost out of stock, another is sitting unsold, and three suppliers are offering different prices, delivery times, and minimum order quantities.
The owner has to connect all these pieces manually: check inventory, understand sales trends, compare suppliers, calculate the actual landed cost, and make sure the purchase stays within budget.
We asked a simple question: What if an AI agent could handle this entire decision-making process, while keeping the business owner in control?
That question inspired OPERANTA.
What We Built
OPERANTA is an AI-powered agentic commerce assistant for small and omnichannel retailers.
Instead of merely answering questions, OPERANTA connects operational intelligence with procurement workflows.
- Unified inventory intelligence: Monitor stock levels and identify products approaching their reorder thresholds.
- Supplier evaluation: Compare supplier reliability, unit prices, freight, lead times, and minimum order quantities.
- AI-powered decisions: Analyse inventory and sales signals to recommend what should be reordered and from whom.
- Financial guardrails: Enforce spending envelopes and per-order limits so recommendations cannot bypass configured business constraints.
- Human-in-the-loop procurement: Route decisions requiring approval to the merchant instead of blindly executing them.
- PayPal integration: Incorporate PayPal into the commerce workflow, demonstrated using a sandbox environment.
The central idea is that AI should not just tell a retailer what is happening. It should help determine what to do next, prepare the action, and respect the merchant's authority.
How We Built It
We built OPERANTA as a full-stack application.
- Frontend: React, TypeScript, and Vite, with AG Grid for interactive inventory management and AG Charts for visual stock analysis.
- Backend: Python and FastAPI for business logic, API orchestration, and agent workflows.
- Database: Supabase and PostgreSQL for persistent application data.
- AI: Gemini for reasoning and operational recommendations.
- Commerce: PayPal integration using sandbox workflows.
- Deployment: Render for hosting the frontend and backend.
The application brings inventory signals, supplier information, spending limits, and procurement decisions into a single operational cockpit.
Challenges We Faced
The biggest challenge was making the AI useful without making it reckless.
A recommendation to purchase inventory is not just another chatbot response. It has financial consequences. We therefore focused on separating AI reasoning from deterministic backend enforcement. Spending limits and order ceilings are enforced by the backend rather than relying on the AI to follow instructions.
Another challenge was integrating several technologies into one working application: frontend, backend, database, AI services, and PayPal. Getting these components to communicate reliably, handling deployment configuration, and creating a coherent end-to-end workflow required considerable debugging and iteration.
We also had to make meaningful product decisions under a tight hackathon deadline: prioritising a working operational workflow over adding features just for the sake of complexity.
What We Learned
Building OPERANTA taught us that agentic AI is not simply about giving a language model access to tools.
A useful agentic system needs context, access to relevant data, clear boundaries, reliable execution logic, and appropriate human oversight.
We learned how to bring AI reasoning together with conventional backend engineering, financial constraints, database-backed workflows, and a real commerce integration.
What's Next?
We envision OPERANTA becoming an intelligent operations layer for independent retailers — one that continuously understands business needs, identifies opportunities, and helps merchants move from insight to action without sacrificing control.
OPERANTA: Intelligence behind every operation.
Built With
- ag-charts
- ag-grid
- agentic-ai
- agentic-commerce
- ai-agents
- fastapi
- gemini
- google-ai
- paypal
- postgresql
- python
- react
- render
- rest
- supabase
- typescript
- vite