On Friday, November 13, one of the world’s largest container shipping companies offered customers a remarkably straightforward solution to a sophisticated technology problem: pick up the telephone. In a customer advisory, A.P. Moller–Maersk warned that the global Microsoft service disruption could delay customer notifications and directed urgent inquiries to local offices by phone. The company emphasized that the disruption originated with Microsoft, not its own internal systems.
That distinction matters. A logistics provider can have functioning operational systems, available transportation capacity, and sophisticated optimization software while experiencing disruptions in the infrastructure that connects it to customers and trading partners. Application availability and operational continuity are not the same thing.
Later that day, Workday experienced a separate outage that interrupted access to enterprise applications. The San Francisco Chronicle reported that a problem discovered during backup-system testing prompted Workday to take some systems offline for investigation. There is no established connection between the two incidents, but together they raise an uncomfortable question: Have supply chain executives engineered resilience into their physical networks while neglecting the digital infrastructure that coordinates them?
Your WMS Is Running. Your Warehouse May Not Be.
A warehouse management system can remain fully operational while the warehouse itself loses important operating capabilities. Conveyors may continue moving cartons, robots may continue transporting inventory, and workers may continue receiving picking assignments. But the facility is a node in a larger network that depends upon communication, information exchange, authorization, and human judgment.
Consider a shipment arriving with a discrepancy. The inventory is physically present, the receiving team is available, and the WMS can process the transaction. Yet supplier correspondence or the authorization required to resolve the exception may be sitting in an inaccessible mailbox. Receiving stalls, inventory remains unavailable for allocation, and downstream fulfillment commitments are jeopardized. The WMS has met its availability objective; the business process has not.
Workday introduces another dimension. Its customer base exceeds 11,500 organizations, including retailers such as Walmart and Target, although the number actually affected by the outage has not been established. Workforce management, time recording, payroll, and financial approvals are not traditionally considered core logistics applications, but their unavailability can complicate labor-intensive operations. With the holiday fulfillment season approaching, the operational significance of these dependencies becomes particularly apparent.
For years, supply chain technology suppliers have emphasized the integration of planning and execution. The next question is whether that integration remains functional when a critical supporting service disappears. A technology portfolio should be evaluated not only on what its applications can accomplish together, but also on what the enterprise can accomplish when some of them are unavailable.
The Economics of Digital Concentration
Supply chain executives understand concentration risk. Few would deliberately source every critical component from a single supplier without evaluating alternate capacity and disruption exposure. Yet the digital equivalent of that sourcing strategy is often celebrated as enterprise standardization.
Consolidating applications, communications, and infrastructure around common platforms offers legitimate advantages: lower administrative overhead, simpler integration, consistent security policies, and a more unified employee experience. The business case can be compelling. What is less frequently quantified is the economic exposure created when multiple operating capabilities become dependent upon the same technology infrastructure.
In systems engineering, common-mode failures occur when multiple components fail because of a shared underlying vulnerability. Two software applications from different suppliers may appear to offer diversification while relying on the same cloud infrastructure, identity provider, or integration service. The enterprise has purchased two applications; it has not necessarily purchased two independent operating capabilities.
The financial consequences extend beyond employee downtime and software subscription credits. A distribution center processing 100,000 orders per day, for example, needs to know how much throughput it can sustain during an outage, whether backlogs can be cleared before carrier cutoffs, and what additional labor or expedited freight recovery will require. The appropriate measure of resilience is not simply system uptime, but the value of the business activity that can be maintained.
The procurement implications are substantial. Software evaluations should examine dependency concentration, the independence of recovery mechanisms, data portability, and the operational feasibility of alternative workflows. The objective is not to duplicate every enterprise application. It is to determine where independent operating capabilities justify their additional cost.
Agentic AI Raises the Stakes
Artificial intelligence will make these dependencies more complex. As intelligent agents increasingly participate in procurement, inventory allocation, transportation planning, and exception management, decisions will depend on a growing ecosystem of enterprise applications, model providers, integration services, and data repositories. Agent-to-Agent communication and the Model Context Protocol can improve interoperability, while retrieval-augmented generation and knowledge graphs can enrich the information available to those agents.
The Microsoft outage offered an early illustration of the other side of that equation. Reuters reported that the disruption also affected Microsoft connectors used by AI services from OpenAI and Anthropic. As organizations build workflows across multiple platforms, the availability of the AI model itself becomes only one element of operational reliability.
Imagine a transportation agent evaluating alternative delivery schedules after losing access to current inventory data. Using previously retrieved information, it might generate an entirely plausible recommendation that no longer reflects operational reality. A second agent could then initiate an inventory allocation based upon that recommendation, propagating the error across the enterprise.
Graceful degradation must therefore become a fundamental AI architecture requirement. Agents need to recognize unavailable dependencies, assess data freshness, distinguish advisory recommendations from authorized transactions, and suspend consequential actions when essential information cannot be verified. The fallback workflow must preserve transaction integrity, not simply keep the AI application responsive.
The most sophisticated autonomous supply chain will not necessarily be the one that makes the greatest number of decisions without human intervention. It may be the one that understands most precisely when autonomous action is no longer appropriate.
Engineering for Operational Continuity
The Workday incident offers a particularly instructive systems engineering lesson. A November 13 incident notice from Florida Atlantic University described an issue involving Workday’s disaster recovery environment that produced inconsistencies in production data. The infrastructure intended to support recovery was itself implicated in the disruption. Redundancy is valuable, but its independence, isolation, and integrity matter as much as its existence.
Supply chain executives should extend their resilience analyses beyond individual software platforms. Recovery time objectives need to reflect how long business processes can tolerate interruption. Recovery point objectives must account for the consequences of lost transaction history, particularly where inventory, orders, shipments, and financial records are synchronized across multiple systems. Restoring an application without reconciling its transactions can create a second disruption during recovery.
Enterprises should also define a minimum viable operating capability. If cloud applications, communications, or AI services become unavailable, how much warehouse throughput and transportation execution can be sustained? Which decisions can be made locally, which require enterprise authorization, and which must wait for reliable information? These questions belong in technology procurement and architecture design, not merely in an IT disaster recovery manual.
A disaster recovery exercise that demonstrates the successful restoration of a server is useful. One that demonstrates the ability to keep shipping customer orders while that server remains unavailable is considerably more valuable. The distinction should inform how enterprises measure resilience and allocate investment.
Maersk’s customer advisory provided a modest but revealing example of this principle. The telephone is hardly a revolutionary logistics technology, but it offered an alternative when the preferred communications infrastructure failed. A supply chain is not resilient because its systems rarely fail. It is resilient because the enterprise can continue to operate when they do.
The post Microsoft, Workday, and the Hidden Fragility of the Digital Supply Chain appeared first on Logistics Viewpoints.