0.1 Series Overview
Agentic Developer: Zero to Advanced builds its whole agent architecture around one use case: X-ray defect detection. This short companion series asks the obvious follow-up question – is any of that actually specific to X-ray images, or is it just a reusable shape? To find out, the identical planning/tool-use/reasoning pattern gets rebuilt around a completely different domain: supply chain.
The business case: late or failed deliveries and demand mismatches cost real money, through expedited shipping, stockouts, or excess inventory. Today a human analyst has to notice something is off and manually dig through order and shipment data to find the cause. An agent can do that investigation automatically: it watches incoming orders and shipments, flags anomalies (unusual delay patterns, a demand spike that doesn’t match forecast), reasons over possible causes (supplier issue, regional disruption, data error), and either recommends an action or drafts an explanation for a human to review – turning a reactive, manual investigation process into a proactive one.
Dataset: the DataCo SMART Supply Chain dataset (Mendeley) – order, shipment, customer, and product data suitable for a demand/delivery anomaly investigation agent. (Alternative considered: the SupplyGraph benchmark, graph-structured supply chain planning data from an FMCG company.)
This is a roadmap post: links go live as each part is written, in order.
0.1.1 Posts in this Series
- Architecture recap – same graph, same agent loop shape as the X-ray series, explained briefly for readers coming in fresh on this series
- Swapping in the supply chain scenario – a demand/delivery anomaly investigation agent built with the identical planning and tool-calling pattern, using the DataCo dataset
- What stayed the same vs. what changed – defect catalog vs. supplier docs, defect severity vs. stockout risk; reflection on architecture portability across domains
Reuse
Citation
@online{goni2026,
author = {Goni, Hasan},
title = {Agentic {Architecture:} {Supply} {Chain}},
date = {2026-09-05},
langid = {en}
}