TECHNOLOGY

Beyond the AI hype: finding the problems worth solving

AI is transforming logistics. But the companies that create the most value will not necessarily be those running the most pilots. They will be the ones that are best at matching technology to real business problems.

Every technology cycle creates a similar temptation: start with the technology and look for places to apply it. Today, AI is powerful enough—and visible enough—that this temptation is particularly strong.

But supply chains do not become more competitive because an organisation has “an AI strategy”. They become more competitive when technology improves a decision, removes friction, increases predictability, reduces cost or creates a service that customers value.

Start with the decision, not the algorithm

In logistics, some of the most valuable problems are not spectacular. They are decisions made thousands of times: which service to book, what lead time to promise, whether a deviation is meaningful, where capacity is likely to be constrained, or which exception deserves attention first.

These are excellent candidates for technology because the scale is high, the data often already exists, and small improvements can compound across large networks.

This is also why I prefer the broader idea of tech-enabled supply chains. AI matters, but it sits alongside automation, data platforms, predictive analytics, robotics, connected assets and other technologies. The useful combination depends on the problem.

Move from descriptive to predictive—and eventually prescriptive

Many supply-chain systems are very good at telling us what has happened. Fewer can reliably explain why it happened, anticipate what is likely to happen next, and help a user decide what to do about it.

That progression—descriptive, diagnostic, predictive, prescriptive—is where much of the real opportunity lies. A lead-time tool, for example, becomes far more valuable when it moves beyond displaying historical averages and can diagnose recurring delay drivers, identify trends and support a routing or planning decision.

The technology may include machine learning. But the difficult work also involves data definitions, process ownership, user needs, integration, adoption and trust.

Trust is part of the product

Supply-chain decisions have consequences. A prediction that nobody understands or trusts has limited operational value. That means explainability, confidence levels, quality of underlying data and clarity about how a recommendation should be used matter as much as model sophistication.

In practice, the strongest technology products combine technical capability with operational credibility. Users need to recognise the logic of the solution and understand where it improves on the process they use today.

Scale changes the innovation question

A successful proof of concept demonstrates that something can work. A successful product demonstrates that it can work repeatedly, within real processes, for enough users to justify the investment.

That shift requires a different mindset. Product ownership, integration, change management, commercial logic and measurable business value become increasingly important. Technology remains essential—but it is no longer the whole story.

That is the standard I believe we should apply to AI in supply chain: not whether it is new, impressive or fashionable, but whether it helps build a supply chain that is more predictable, efficient, resilient and competitive.