This article was originally published on LinkedIn.
In a contested fight, logistics isn’t a support function.
It’s a target.
In large-scale combat operations, the adversary doesn’t just attack fuel farms and convoys. They attack timing. They attack predictability. They attack the decision cycles that move sustainment forward.
That’s why I believe one of the most important challenges in contested logistics is not technological.
It’s architectural.
Precision Sustainment is not just a data problem
Precision sustainment sounds like a forecasting and data problem.
But it’s really a decision-rights problem.
Who is authorized to trigger replenishment? Based on what signals? With what guardrails when networks degrade? And how do we prevent well-intended local decisions from creating system-wide volatility?
AI can accelerate decisions.
But it cannot fix a poorly designed decision system.
A lesson from commercial supply chains
Decades ago, large retailers and manufacturers faced a similar structural problem:
- Volatile downstream demand
- Spikes caused by panic ordering
- Poor transportation utilization
- Excess “just in case” inventory
- Expensive emergency shipments
The breakthrough was Vendor-Managed Inventory (VMI).
The core idea was simple but powerful:
Instead of every downstream node independently placing orders based on partial information, replenishment responsibility shifted upstream — to the party with broader visibility and the ability to coordinate across the network.
Consumption signals drove replenishment. Decision rights were clarified. Replenishment became coordinated instead of reactive.
Inventory dropped. Service improved. Variability declined.
The important insight wasn’t technological.
It was architectural.
Why this matters for contested logistics
Now translate that structure to multi-domain operations.
Forward units see consumption first. The theater sustainment enterprise sees capacity and constraints across the force.
When communications degrade and formations disperse, uncoordinated replenishment requests create volatility:
- Duplicate orders
- Inflated safety stock
- Distribution surges
- Predictable resupply patterns
- Overloaded lift assets
In a commercial setting, that wastes money.
In a contested fight, it increases risk.
Imagine a dispersed brigade combat team operating across 200 kilometers of contested terrain. After an unplanned maneuver, fuel consumption spikes. Two battalions independently submit emergency requests. Network connectivity is intermittent. The theater node sees incomplete signals. Aviation and ground lift are limited.
Emergency movements get prioritized.
Signature increases.
Capacity tightens.
The system reacts.
Now imagine a different model.
Consumption signals feed into an enterprise replenishment cell operating under predefined rules. Allocation decisions are coordinated across battalions. Distribution is smoothed. Guardrails prevent overlapping emergency requests unless thresholds are exceeded.
The system absorbs volatility instead of amplifying it.
That is not primarily an AI problem.
That is decision architecture.
Introducing “Contested VMI”
Retail VMI cannot simply be copied into a warfighting environment.
But the core mechanisms are highly relevant:
1. Demand uncertainty linked directly to replenishment Not just dashboards — decisions.
2. Reduced uncertainty through coordinated policy Eliminating panic-driven ordering and smoothing flows.
3. Intentional inventory positioning Placing inventory based on readiness risk, not just storage convenience.
4. Coordinated replenishment across dispersed nodes Maximizing limited transportation capacity.
5. Capacity smoothing as a survivability advantage Reducing predictable surges that create targeting windows.
In a contested environment, this becomes something different.
Call it Contested VMI.
It would require:
- Operating with intermittent connectivity
- Predefined replenishment guardrails at the edge
- Mission-based service definitions (readiness, uptime, continuity of care)
- Risk-aware allocation logic
- Explicit doctrine for replenishment authority
Most importantly, it requires clarity about who decides what — and under what conditions.
AI is an accelerator, not an architect
We often to algorithms.
Better prediction. Faster optimization. More data fusion.
Those matter.
But if replenishment authority remains fragmented, if every node reacts independently under stress, and if policy guardrails are unclear, AI will simply accelerate volatility.
Precision Sustainment is not only about machine speed.
It is about disciplined, coordinated replenishment under adversarial pressure.
The architecture has to come first.
A practical way forward
This does not require rewriting doctrine overnight.
It could start with a pilot:
- Select a high-impact repair part family or medical category
- Instrument consumption at the point of use
- Establish replenishment decision rights at an enterprise cell
- Define edge guardrails for degraded network conditions
- Measure readiness outcomes, not just inventory levels
The goal would not be to reduce inventory for its own sake.
It would be to reduce volatility, improve readiness stability, and smooth sustainment signatures under contested conditions.
If the pilot demonstrates fewer emergency movements, fewer stockouts at the point of need, and more stable distribution rhythms, then the architecture works.
The bigger point
Contested logistics will never be easy.
But we do not need to invent every answer from scratch.
Some of the most powerful advances in supply chain performance over the last thirty years were not about better forecasting.
They were about better decision design.
The Army does not primarily need better logistics AI.
It needs better decision architecture.
If we redesign replenishment authority, align visibility with responsibility, and build guardrails that function even when networks fail, AI can amplify that system.
Without that architecture, AI will only make a reactive system faster.
And faster chaos is still chaos.
Matthew Waller, M. Eric Johnson, Tom Davis. “Vendor Managed Inventory in the Retail Supply Chain,” Journal of Business Logistics, 20(1), (1999), 183-204.
