Lora Cecere warns against the AI spin cycle in supply chain planning

Lora Cecere published a Supply Chain Shaman article arguing that supply chain technology is entering an AI “spin cycle”: AI is being applied across product portfolios, but without enough clarity on definitions, interoperability or work redesign. The article reflects on recent events hosted by Kinaxis, project44 and Optilogic, noting product progress but questioning whether the market is addressing the deeper operating problem.

The strongest point is that supply chain leaders may be asking the wrong question. Cecere argues that the goal should not simply be integration between planning and network design systems. Integration creates data exchange. Interoperability enables scalable collaboration across organizations, technologies and decision contexts. That distinction matters because AI value depends on common meaning, synchronized data and harmonized decision logic, not just APIs.

The article’s practical agenda starts with a planning master data layer powered by Semantic Reconciliation. Instead of treating data quality as a complaint, the proposal is to use machine learning, agentic workflows and rule-based ontological frameworks to reconcile meanings across systems. For supply chain planning, this is a governance issue: if demand, lead time, inventory, cost and service definitions are inconsistent, AI will scale confusion rather than intelligence.

Cecere also argues for changing how network design is used. Rather than leaving network design analysts in project-based or functional roles, she recommends using network design systemically and pairing it with finance to build a “plan of plans.” In this model, strategic network scenarios flow into tactical planning and S&OP, allowing plans to learn from each other rather than simply consuming more data.

The article also highlights lead time as a planning variable, not a static master-data field. Many organizations invest in logistics visibility to predict on-time delivery, but do not use those signals to inform planning decisions. Cecere’s point is that lead-time shifts should change safety stock, buying plans and customer go-to-market decisions. This is where visibility data from providers such as project44 or FourKites becomes more valuable when connected to a governed Planning Master Data layer.

The closing message is clear: do not sprinkle AI over existing planning taxonomies and call it transformation. The objective is not faster decisions by default. It is better outcomes, clearer value definitions and a supply chain operating model that can use AI to improve work rather than amplify existing fragmentation.

Dataleo angle

This is a strong Radar insight because it pushes Supply Chain AI back toward decision architecture. The practical question is not whether a vendor has an AI narrative, but whether AI improves the decisions that matter: make, source, deliver, buy, allocate, buffer, expedite or redesign the network.

For planning leaders, the governance checklist should be explicit: what data meanings are reconciled, who owns the planning master data layer, how lead time becomes variable, how network-design scenarios connect to S&OP, and where agentic workflows are allowed to automate decisions. Without this discipline, AI risks becoming a faster version of the same fragmented planning model.