Supermarket Bot Manager Automation

use case

Inspiration

Indonesia's regional supermarket industry is quietly drowning in operational chaos. 73% of regional chains still manage supplier orders manually, juggling 12–20 suppliers per store with zero centralized price comparison — and losing Rp 2–8 million monthly from overstock and expired goods alone. Managers are overwhelmed: tracking hundreds of SKUs across branches, generating P&L reports days after the fact, and negotiating with 15+ suppliers individually with no benchmark and no automation.

We built Supermarket Bot Manager Automation because we believed this problem deserved a real, agentic solution — not just a dashboard, but an AI that works like a full back-office team. Our goal was clear: reduce operational overwork through AI-powered automation for stock tracking, bookkeeping, and bill payments, while increasing business efficiency and profit by preventing stock shortages and automatically recommending the best supplier prices.

What it does

Supermarket Bot Manager Automation is a multi-agent AI system that acts as an autonomous back-office for supermarket managers — all accessible via WhatsApp.

A manager simply texts: "Gimana keadaan toko hari ini?" ("How's the store today?") and the system goes to work through 5 specialized AI agents:

Supervisor Agent — Parses the manager's intent and routes tasks to the right agent
Stock & Forecaster Agent — Monitors inventory levels and forecasts demand using real-time weather data from BMKG API (e.g., flagging that rice and gallon water are critical by evening)
Buyer Agent — Finds the best supplier prices automatically via Tokopedia & Shopee API and sends pre-orders to suppliers
Cashier Agent — Handles customer orders and generates QRIS payment links via DOKU API
Bookkeeper Agent — Compiles daily financial reports and logs all transactions through DOKU

At the end of the workflow, the Supervisor aggregates everything and sends the manager one clean, consolidated reply — covering stock status, purchase orders, revenue, and profit for the day.

How we built it

The core of the system is built on OpenClaw as our LLM core and framework. Our dominant programming language was JavaScript, with the two of us splitting responsibilities between backend and frontend.

Key integrations include:

WhatsApp Business API — primary interface for manager communication
Tokopedia & Shopee API — real-time market price data for procurement
DOKU API — payment processing (QRIS) and transaction bookkeeping
BMKG API — weather and environmental forecasting to drive stock predictions

The multi-agent architecture was designed so each agent operates independently on its domain, but reports back through the Supervisor — keeping the manager's experience simple while the complexity runs in the background.

Challenges we ran into

Our biggest constraint was time — we had only 10 hours to build a system with significant architectural complexity. Finding and integrating the right APIs consumed a large chunk of that window, as each integration came with its own documentation, auth flows, and edge cases.

Limited prior experience with multi-agent at this scale also added to the challenge. There were moments where the system design had to be rethought on the fly to keep everything connected and running within the deadline.

Accomplishments that we're proud of

The moment the AI agent ran end-to-end — from a WhatsApp message to a consolidated store report — exactly as we had designed it, was deeply rewarding. Hours of architecture planning, integration work, and debugging compressed into a working system that genuinely solves a real problem for real people.

Seeing all 5 agents coordinate, each doing their job and handing off to the next, made every difficult hour worth it.

What we learned

This project taught us that the complexity you anticipate is only a fraction of what you actually encounter. Challenges we expected appeared in different forms, and new ones emerged along the way — from API inconsistencies to agent coordination edge cases.

More importantly, we learned that pushing through those challenges is exactly what makes you grow as an AI programmer. We came out of this hackathon with a sharper understanding of agentic system design, real-world API integration, and how to ship fast under pressure.

What's next for Supermarket Bot Manager Automation

The immediate next step is developing a Marketer Agent — a sixth agent that autonomously runs promotions based on overstock levels and upcoming expiry dates. Instead of letting slow-moving inventory become a loss, the Marketer Agent will trigger targeted discount campaigns automatically, turning a cost center into a revenue recovery mechanism.

Long term, we see this platform expanding to serve the thousands of regional supermarket chains across Indonesia that are still operating without any intelligent automation layer.