Rajeev Gangwar
← Back to Case Studies
Agentic AI

Agentic Supply Planning POC

Built a multi-agent workflow that runs demand-supply balancing and inventory optimization with human-in-the-loop guardrails.

Agentic AIBPMNMCPTemporalFastAPI
The Challenge

Demand forecasting and inventory optimization are routine decisions that still eat planner time — weekly or monthly cycles where data is pulled, models are run, safety stocks are recalibrated, and recommendations are reviewed. The analytical work has been solved for decades; the bottleneck is the orchestration around it.

The Approach
1

Built a BPMN-driven workflow that coordinates specialized agents through demand-supply balancing: agents pull history, invoke a FastAPI forecasting microservice (time-series + ML), pass results to an inventory-optimization service (safety stock, EOQ, ABC), generate replenishment recommendations, and surface them as tasks for human review.

2

MCP-based integrations connect to ERP systems for real data.

3

Temporal handles long-running workflow state.

The Outcome

A working POC where the agent loop runs the full plan-review-adjust cycle against a live ERP with demo data. Planners see recommendations, not raw numbers — and the system keeps a full audit trail of which agent did what, when, and why.

The Lesson

Agentic AI shines in the operational middle — not the strategic decisions (humans) and not the routine transactions (existing ERP) — but the orchestration between them, where well-bounded agents under clear guardrails can handle decisions that repeat.

RG
Ask Rajeev's AI
Online now