Resilient Supply Chain Podcast: Resilience Fails in the Workarounds BetweenSystems
Supply chain disruption is usually associated with ports, suppliers, transport capacity or geopolitical shocks. Episode 143 of the Resilient Supply Chain Podcast, hosted by Tom Raftery, examines a less visible source of operational risk: the spreadsheets, document hand-offs and improvised processes connecting formal systems. Guest Patrick Van Hull, now an Agentic Operations Consultant at Salesforce and an Industry Consultant at Tungsten Automation when the conversation was recorded, argues that seemingly minor workflow weaknesses can cascade into delays, lost productivity and production disruption. For supply chain leaders investing heavily in automation and AI, the central question is whether technology is improving the process or simply accelerating an existing workaround. The full episode is available at www.resilientsupplychainpodcast.com
When a Workaround Becomes Infrastructure
Many supply chain processes are held together by temporary fixes that ceased being temporary years ago. Van Hull describes ERP exports moved into Excel, planning data joined through VLOOKUPs, and information exchanged manually because formal systems cannot keep pace with operational requirements. The problem is not necessarily that such workarounds fail immediately. Their danger is that they often succeed.
Van Hull recalls a product-launch process in which Excel files were exchanged directly between individuals without structured checks. Because the approach worked initially, it was repeated until it effectively became part of the business process. Asked when such workarounds become business risks, his answer is blunt: "The day they're instituted."
The consequences can move quickly from information management into physical operations. An inbound supply chain may require more than 30 separate documents, from purchase orders and bills of lading to shipment notifications, goods receipts and invoices. A missing hazardous-material declaration, for example, can leave an entire container waiting even when every other part of the shipment is ready.
Automation Can Amplify the Wrong Process
The episode challenges the assumption that AI or automation necessarily resolves this fragility. Van Hull argues that organisations often begin by automating the process they already have rather than asking whether that process should exist in its current form.
"If you're trying to automate the workaround, you're just accelerating it," he says. "You're accelerating the inefficiency."
That distinction matters as companies pursue document intelligence, robotic process automation and agentic AI. Automating one well-defined step may improve local performance while leaving upstream and downstream dependencies unchanged. Accounts payable, supplier onboarding, logistics receiving and inventory may each become more efficient individually while the overall flow remains fragmented.
The governance implication is that automation programmes need to be assessed end to end rather than function by function.
Human Judgement Still Defines the Boundary
Van Hull proposes a practical dividing line between automation and human decision-making. High-repetition, low-variability activities are natural candidates for rules, automation and machine assistance. High-stakes, high-variability decisions require considerably more human involvement.
That places accountability at the centre of AI adoption. An AI system may narrow options or recommend an action, but executives still need to decide what outcome is being pursued, which trade-offs are acceptable and who remains responsible for the final decision.
The same principle applies to process design. Frontline teams often know the exception paths that formal process maps miss because they experience what happens when pallets are waiting, documents are incomplete or systems stop connecting. Excluding that operational knowledge risks automating an idealised version of the process rather than the process that actually exists.
Incentives Can Preserve Inefficiency
Automation also exposes an organisational problem. Employees whose performance is measured by transaction volume may see successful automation as a threat rather than an improvement. Van Hull describes teams evaluated by the number of forecast items they manually adjusted. Once a system could perform much of that work automatically, the obvious question became whether fewer manual interventions made those employees appear less valuable.
That creates a governance challenge extending beyond technology. Redesigning workflows may also require redesigning performance measures, accountability and definitions of contribution.
For supply chain leaders, the broader lesson is that operational resilience depends as much on the integrity of everyday processes as on preparation for major external disruption. AI and automation can reduce repetitive work and improve consistency, but they cannot compensate indefinitely for unclear ownership, disconnected functions or processes that nobody has revisited because they still appear to work. The more interconnected supply chains become, the greater the need to identify these hidden dependencies before they become embedded in the next generation of automated operations.
