Supply chains are investing heavily in AI, predictive tools and digital platforms, but technology alone does not determine whether those investments improve performance. Data quality, decision rights and workforce adoption can decide whether a new capability changes operations or simply adds another layer of technology.
At the Supply Chain Management Summit at Supply Chain LIVE on September 9, representatives from Mars, Royal Papworth Hospital NHS Foundation Trust, Pladis and Kimberly-Clark discussed what determines whether AI investment delivers measurable value across supply chain operations.
AI Value Starts With the Operating Model
Supply chain performance increasingly depends on decisions made across inventory, capacity, service, resilience and investment. That gives the function greater influence over business performance, but the definition of a successful supply chain can vary considerably between operating environments.
Healthcare provides an especially clear example. Kenny Otto, Head of Supply Chain Management at Royal Papworth Hospital NHS Foundation Trust, said product availability can directly affect whether clinicians are able to perform procedures.
“Without having products at the side of the consultants when they’re carrying out their procedures, the supply chain is not functional because then the failure is at that bedside when a patient cannot get the treatment they need,” Otto said.
In consumer businesses, the consequences may instead appear through service, inventory, margins or the ability to respond to changes in demand. Abdelaziz Salah, VP Supply Chain International Family Care & Professional at Kimberly-Clark, argued that there is no universal model for a successful supply chain and that performance needs to align with the wider business strategy.
James Rosengren, Global Supply Chain, Value Creation and Sustainability at Mars, went further, describing supply chain as a source of competitive advantage when it participates directly in business decisions.
“My big belief is the supply chain is a competitive advantage,” Rosengren said. “It’s really important that rather than seeing it as a service function within the business, if you set it up to say you have a seat at the table, the supply chain is a key strategy enabler and decision maker.”
The same principle matters when companies invest in AI. A forecasting platform, predictive model or visibility system has limited impact if the organization has not established who acts on its recommendations, how quickly decisions can be changed and which operating outcome the investment is intended to improve.
AI Investment Can Outrun Operational Readiness
Digital capabilities can be deployed faster than the processes and behaviors required to use them effectively.
Suleyman Nezih Kocaman, Senior Director, Global Supply Chain at Pladis, said the bigger obstacle is often not the technology.
“I think the biggest challenge is not the technology itself,” Kocaman said. “It’s more of a people and decision making problem, not the technology problem at the moment. Technology for me is more an enabler providing you the visibility that you need.”
Better signals do not automatically produce better decisions. Teams still need confidence in the recommendation, authority to act and enough operational flexibility to change the plan.
Data quality can create another constraint. Rosengren emphasized the importance of understanding what a platform can realistically deliver and whether the organization has the master data and information required to support it.
Salah also pointed to master data and change management as essential parts of implementation. Poor inputs weaken the insights generated by AI, while inadequate adoption can leave technically capable systems sitting outside normal operating decisions.
Companies can therefore continue investing in new platforms without fully capturing the return from existing ones. Adding another forecasting model or AI interface does little to improve performance when decision processes, data ownership or adoption remain unresolved.
Predictive AI Raises the Bar for Execution
Predictive systems create a different operational test because identifying a problem earlier only matters when the organization can still do something about it.
Salah described applications in which machines can identify deviations and adjust energy and water consumption to help keep production within required parameters. In that case, detection is connected directly with an operational response.
Otto pointed to predictive AI as a way to anticipate risks including future cyber incidents and stockouts.
The distinction matters. Predicting a stockout does not prevent one if purchasing cannot change an order, inventory cannot be repositioned, an alternative product cannot be approved or a supplier cannot respond within the remaining window.
Manufacturing and logistics face the same issue. Earlier detection of a production constraint, transportation delay or inventory imbalance increases the time available to respond, but the financial benefit depends on whether viable alternatives remain available.
As predictive capabilities improve, this ability to act may become a more important source of return than prediction accuracy alone.
The Better AI Metric Is Decision Time
The harder measure of AI performance may not be how many processes use the technology or how quickly a model identifies an exception. It is the time between identifying a problem and changing an operational decision.
A system that predicts a stockout days earlier creates little value if purchasing approvals, supplier constraints or inventory policies consume that advantage. The same applies to production deviations, transportation disruption and capacity shortages. Earlier information becomes economically useful when there is still enough time and flexibility to act.
Tracking how quickly an AI-generated signal becomes an operational decision would give companies a more useful measure of whether the technology is changing performance or simply increasing the volume and speed of information.
