Supply chain feedback loops have always been important, but problematic to master. Change could now be at hand. This article is a continuation of our first installment, A Short History of Supply Chain Feedback Loops. These stories are part of our STACKED series.

For the first time since Forrester studied that Kentucky appliance factory, the nature of the supply chain signal itself is changing, not just the speed at which it travels.
Sensor technology and the Internet of Things are reducing latency across supply chains in ways that would have been practically impossible a decade ago. Pallets report their own departure. Retail shelves report their own depletion. Containers crossing the Pacific report position, temperature, and estimated arrival continuously. Instead of batched records of transactions that have already occurred, supply chain managers are beginning to see something closer to a continuous stream of operational reality as it unfolds.
Artificial intelligence is simultaneously expanding the range of signals that can inform decisions. Classical demand forecasting worked from order history, looking backward at what was purchased last week or last quarter and projecting forward. Modern demand sensing systems can ingest weather forecasts, social media sentiment, web search trends, promotional activity, point-of-sale data updated by the hour, and macroeconomic indicators, integrating sources that no human forecaster could synthesize at speed. In practical terms, companies are beginning to detect shifts in consumer demand before those shifts fully appear in the purchase orders moving upstream through the supply chain.
Digital twins complete the picture. Forrester’s original contribution in Industrial Dynamics was to build a dynamic simulation of a supply chain, a model that could represent the system’s feedback structure and trace the consequences of decisions through time. The digital twin pursues the same objective, but is fed by live operational data and runs continuously. It is Forrester’s method at operational scale, in real time.
Together, these three forces address, for the first time at commercial scale, the specific conditions Forrester identified as the root cause of demand amplification: information delay, signal distortion, and structural opacity. The case for optimism is real. Inventory buffers shrink when uncertainty shrinks. Order batching becomes unnecessary when replenishment is continuous. Safety stock strategies evolve when lead time visibility improves. These are genuine structural improvements, not incremental efficiency gains.
But there is another side to this story, and it deserves equal attention.
The Problem With Synchronized Signals
The classical bullwhip effect was, paradoxically, partly contained by the very disconnection that caused it. Because firms operated on independent, delayed signals, their errors were staggered. One company’s over-order was offset by another’s under-order. The oscillations were severe, but they were not synchronized across entire industries simultaneously.
AI adoption may change that structural condition in ways the field has not yet seriously examined.
Consider what happens as a significant share of large firms in a given sector begin running similar machine learning models, trained on overlapping datasets, drawing on the same real-time signals, and updated by the same external data sources. Those firms, which previously operated as independent decision-makers, begin to behave as correlated ones. The same signal reaches many firms’ models at roughly the same time. Models trained on similar data and optimized for similar objectives tend to generate similar responses. And when those responses are executed with a high degree of automation, the human deliberation that previously introduced variation and delay is reduced or removed.
The concern, which is structural and logical rather than empirically established, is that this dynamic could produce synchronized ordering behavior across an industry, generated not by actual shifts in demand but by the simultaneous response of correlated AI systems to a shared signal. If that reasoning holds, the bullwhip effect would not have been eliminated by AI adoption. It would have been rewired.
A Parallel Worth Considering
The concern is not without precedent, though the parallel comes from a different industry and should be treated as illustrative rather than definitive.
On May 6, 2010, the Dow Jones Industrial Average dropped nearly 1,000 points in minutes before largely recovering within the same session. The event became known as the Flash Crash. Investigation revealed it began with a single large automated sell order that triggered a chain reaction among high-frequency trading algorithms, each individually rational, each responding to the same cascading signal, collectively amplifying a manageable disturbance into a market-wide shock. The underlying economic reality had not changed. The system’s own automated response mechanisms had manufactured the crisis.
The structural echo for supply chains is worth considering carefully, while acknowledging that the analogy has real limits. Financial markets operate in milliseconds on liquid assets. Supply chains move physical goods through systems that take weeks or months to respond. That slower physical reality could actually make certain forms of AI-driven overreaction more consequential, not less, because corrections cannot occur instantaneously once production and procurement decisions are already in motion.
A synchronized AI-driven demand spike in a physical goods sector would also be harder to identify as artificial than a stock market anomaly. Models generating the response would interpret corroborating signals as confirmation. Inventory would genuinely be depleted. Prices would genuinely rise. Suppliers would genuinely struggle. Every observable indicator would point to a real shortage, because the algorithmic response would have created one.
This paper does not claim this outcome is inevitable or that it has yet occurred at scale in any supply chain sector. What it does suggest is that the structural conditions for it are being created, and that the question deserves serious attention before the answer becomes visible through a crisis.
The Question Supply Chain Leaders Should Be Asking
Forrester’s deepest insight was not about inventory or forecasting. It was about the relationship between system structure and system behavior. Complex systems cannot be improved by optimizing individual components in isolation. Behavior emerges from structure: the feedback loops, the time delays, the decision rules connecting the parts.
The firms currently deploying AI demand sensing, IoT networks, and digital twins are, almost universally, doing so at the level of the individual firm. They are asking how these tools improve their forecasting accuracy, their inventory efficiency, their service levels. These are legitimate questions. They are also incomplete ones.
The structural question worth adding is this: what happens to the feedback loops of the broader supply chain system when many firms deploy these tools simultaneously? How does correlated AI behavior change the dynamics of the system as a whole?
In practical terms, this means supply chain leaders should be stress-testing their AI systems not just against historical demand patterns but against scenarios of synchronized industry response. It means asking where the human decision points are that can interrupt an automated response before it contributes to a system-level amplification event. It means treating AI-driven demand amplification as a scenario worth planning for, alongside the more familiar risks of supplier disruption, logistics failure, and demand shocks.
After the 2010 Flash Crash, financial regulators introduced market-wide circuit breakers, automatic pauses triggered when price movements exceeded defined thresholds, specifically to create the intervention window that correlated algorithmic behavior had eliminated. Supply chains do not have a regulator in the same sense. That means the circuit breaker function needs to be built into organizational governance rather than imposed externally.
The Structure, Again
Nearly seventy years ago, Forrester argued that supply chain instability was not caused by bad actors or poor decisions. It was caused by the structure. Rational local decisions, made independently by well-intentioned people with incomplete information, combined to produce irrational system-wide outcomes. The solution, he argued, was not to find someone to blame but to understand the feedback loops generating the behavior.
The technology arriving now is the most powerful set of tools the field has ever had for acting on that insight. Real-time visibility, AI-driven sensing, and dynamic system modeling are genuinely transformative. The organizations that deploy them thoughtfully will operate with meaningfully lower costs and higher resilience than those that do not.
But thoughtful deployment means asking the structural questions, not just the operational ones. Not only how accurate our forecast is, but also what feedback loops our forecasting behavior is participating in. Not only how fast can our systems respond, but what happens when all of our competitors’ systems respond at the same speed in the same direction.
Forrester spent his career arguing that you cannot fix a system by managing its outputs. You have to understand its structure. The tools have finally caught up with his diagnosis. The thinking now needs to catch up with the tools.
This article is based on our white paper: FROM FORRESTER TO FEEDBACK: How AI Is Finally Solving the Bullwhip Problem
Rick LeBlanc, EMBA, is the founder and editor of Reusable Packaging News (subscribe to the free newsletter) and editor of Western Pallet Magazine. A supply chain journalist with more than 30 years of experience, he specializes in pallets, reusable packaging, material handling and related operations. Rick is co-author of Pallets: A North American Perspective and Pallets & Progress: A Collected History of Pallets and Palletized Handling 1922–1945. He is also an advisory board member of the Virginia Tech Center for Packaging and Unit Load Design. Read Rick’s full bio.


