AI supply chain diagnostic: how a mid-market company identifies 5-10M EUR in value leaks

By combining structured business interviews, ERP data extraction, and the construction of a flow Knowledge Graph, a diagnostic spanning a few weeks identified 5 major value leaks in a €115M revenue mid-market company, including €2–4M recoverable as quick wins within 6 months.

The context: a mid-market company under pressure

A French mid-market company, industrial manufacturing, €115M revenue, 2,000 employees. Complex supply chain: long lead times, multiple suppliers, mix of standard products (70–80%) and custom (20–30%).

The SAP S/4HANA ERP was recently migrated. Modules are in place (FI/CO, SD, MM, PP, PM). But the processes didn't follow. Planning parameters were not recalibrated. The sector market is in crisis: −17.6% registrations in 2024. Every working capital point optimised counts.

How did the diagnostic unfold?

The diagnostic followed the first two steps of the ICPC framework: Identify and Frame.

Structured business interviews. 4 interviews with the order management, purchasing, production, and logistics teams, guided by the GenieFactory platform. Objective: capture the real work, not the declared processes. The gap between the two is systematically the source of value leaks.

SAP data extraction and cross-referencing. 8 cleaned and modelled SAP exports: items, stock, movements, orders, manufacturing orders, BOMs, accounting. A "planning-ready" pivot model built from the real data.

Knowledge Graph construction. The 4 interviews are transformed into 4 process KGs, merged into a unified KG. This graph links the flows, from quote to production to delivery, with real business constraints, not the ERP's theoretical processes.

Enriched scorecard. 16 KPIs calculated across the PLAN→BUY→MAKE→STOCK→DELIVER chain. Each KPI is enriched with root causes extracted from the Knowledge Graph. The scorecard moves from "you have a problem" to "here is why and how to fix it."

What value leaks were identified?

Five major value leaks, with quantified scenarios for each:

1. Structural raw material overstock. Theoretical lead times in the ERP do not match actual supplier lead times. Result: the company holds raw materials longer than necessary. Impact: €3–5M in tied-up working capital, out of €24.3M total raw materials.

2. Excessive work-in-progress (WIP). Manufacturing orders are launched without all components available ("no full kit"). Bottlenecks multiply. Impact: €1–3M.

3. MRP nervousness. Planning parameters (safety stocks, lead times, lot sizes) have not been recalibrated since the SAP migration. Safety stocks are "round numbers unchanged for 2 years." Result: a flood of contradictory "Reschedule In/Out" messages that nobody processes. High indirect impact.

4. Compensatory premium transport. Express deliveries to compensate for structural delays. The emergency mode has become standard operating procedure. Impact: €200–500K/year.

5. Dead and slow-moving stock. Sleeping capital. Impact: €500K–1M.

Estimated total impact: €5–10M, of which €2–4M recoverable as quick wins within 3–6 months.

Why does AI make a difference in this type of diagnostic?

Three concrete reasons:

Data × business knowledge cross-referencing. A traditional consultant analyses either the ERP data or the business interviews. The Knowledge Graph cross-references both. An inconsistency between the lead time declared by purchasing and the actual lead time in the SAP data appears automatically.

Speed. What would have taken months in traditional mode (on-site audits, manual extraction, consolidation) is accomplished in weeks. AI agents do the cross-referencing work; the expert validates.

Capitalisation. Everything learned, business ontologies, diagnostic patterns, scoring rules, is encoded in the Knowledge Graph. The next industrial diagnostic (same sector or adjacent sector) starts from this base. That's the "C" of Capitalise in the ICPC framework.

Is this type of diagnostic applicable to my company?

If you're an industrial mid-market company with an ERP (SAP, Sage, Oracle) and a complex supply chain, yes. The model is designed to be reproducible. The technical building blocks (ERP connector, extraction pipeline, Knowledge Graph engine) are the same, what changes is the business ontology for your sector.

GenieFactory works with specialised vertical partners who provide sector expertise. The platform provides the AI infrastructure. The client retains ownership of their data and their Knowledge Graph.

Launch a supply chain diagnostic →

---

Related article: What is agentic transformation?