Procurement Agent Coherence Score (PACS)

GMI-MROSC-006MRO Supply ChainIntegration-RequiredPattern O-6: Autonomous Coherence

Definition

The percent of decisions across multiple plant-level AI procurement and replenishment agents that align with network-level optimization objectives rather than conflicting, synchronizing destructively, or amplifying variance.

The Measurement Gap

Plant A's reorder agent places a 500-unit PO on Vendor X for bearings. Within the same 48-hour window, Plant B's agent places a 400-unit PO and Plant C's agent places a 350-unit PO: all on the same vendor for similar items. Individually, each decision is rational. Collectively, they hit Vendor X with 1,250 units of demand in a window when the vendor's capacity supports 800. The vendor throttles delivery, all three plants experience delay, and the expediting cascade costs more than the inventory savings the agents were optimizing for. No plant-level metric sees the problem because each plant's metrics show a properly triggered, properly sized PO. The coordination failure is visible only at the network level, and no standard MRO supply chain metric operates at the network level. Inventory turns, fill rate, PO cycle time, and vendor on-time delivery are all computed plant-by-plant. Synchronized POs, parallel safety stock changes, and AI-amplified demand spikes at shared vendors go undetected until vendors respond with capacity constraints or backorders. The failure mode is not that individual agents are wrong: it is that their collective behavior produces outcomes no individual agent can see.

Decision Supported:

Whether to centralize replenishment decisions, install a network-level arbiter agent, or constrain plant-level agent autonomy on shared vendors and shared capacity.

Formula

Plain language: When two or more plant agents make decisions on the same vendor, SKU, or capacity within a coordination window, do those decisions align with what a network-level optimization would recommend? Count the fraction that align.

Notation:

PACS = aligned_decision_pairs / total_interacting_decision_pairs

Variables:

Variable Definition Unit
aligned_decision_pairs Decision pairs within tolerance of network-optimal recommendation Count
total_interacting_decision_pairs Decisions touching shared resource within a coordination window (e.g., same vendor + 48-hour window) Count

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Monthly (rolling coordination window)
Category
C - Network Coordination & Bullwhip
← Auto-Reorder Bullwhip Amplification (ARBA)Criticality-Optimization Divergence Index (CODI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment