Stop Selling AI as Magic. Start Using It Where It Actually Helps
A practical view of AI in the supply chain, from the shop floor to the carpeted floor
I get increasingly frustrated by some of the unsophisticated marketing around AI in supply chain. Too often, the same examples are repeated as if they are new discoveries: optimise the machines; improve the forecast; monitor equipment to reduce downtime. Yes, these are all valid use cases, but they are not quite the revolutionary breakthrough they are often presented to be.
Forecast optimisation has been delivered and refined over many years. AI embedded in tools such as Microsoft Demand Planning can provide additional insight, and the underlying mathematics has progressively improved. Supply chain optimisation has also been supported by tools such as Preactor and partners such as Kudos. Predictive maintenance is also valuable, but in many cases, the people best placed to learn from machine failure patterns are the original equipment manufacturers, because they have the breadth of data across installed assets.
So, the question is not whether these use cases matter. They do. The question is whether they are the best place for traditional supply chain businesses to start if they want AI to make a practical difference now.
Use AI for what it is good at
I am a great believer in identifying talent and deploying it where it creates the most value. You would not put your best sprinter in the shot-put event. The same principle applies to AI. Before we ask where to deploy it, we should first ask what it is actually good at.
Large language models are excellent at interpreting information and generating responses when given the right guidance.
Analytics engines are good at finding trends, insights, probabilities and exceptions across large data sets.
Machine learning can improve processes over time by learning from outcomes and moving towards better patterns of performance.
These capabilities are huge, but they are not magic. They are most useful when they are applied to the right work: the repetitive, information-heavy, communication-heavy and decision-making activities that consume time without creating proportional value.
The carpeted floor is the new factory floor
This builds on a theme I have written about before: supply chain automation has a blind spot. For the last 40 years, we have been very good at walking the shop floor. We have studied waste, movement, changeovers, downtime, inventory, quality and throughput. Lean, Six Sigma, Kaizen and The Toyota Way have all taught us to look carefully at visible operational waste.
But have we walked the office with the same intensity? In many supply chains, some of the biggest piles of waste are not sitting next to a machine. They are sitting in inboxes, spreadsheets, trackers, approval loops, supplier chases, manually prepared documents and repeated requests for information. They are hidden because the work looks like normal business.
This was a central point in my article, moving from AI hype to business value. AI benefit is not created by isolated pilots or clever demonstrations. It is created when technology is applied to real business problems, supported by the right data, embedded in the operating model and adopted by the people doing the work every day.
What the AI Supply Chain Day showed us
At our recent AI Supply Chain Day with Microsoft, we deliberately started with work rather than technology. The aim was not to run a speculative AI showcase. It was to ask supply chain professionals where work gets stuck, where people are wasting time, and where AI could remove friction from real processes.
The outputs were refreshingly practical. Across familiar process areas such as Inquiry to Cash, Purchase to Pay, Plan to Produce and Design to Operate, the discussion kept returning to administrative load: supplier chasing, document creation, clarification loops, spreadsheet updates, status reporting, approval preparation, data reconciliation and communication between teams.
Some of the strongest ideas were not about replacing people. They were about protecting the time of good people. In procurement, agent-supported supplier selection could help draft specifications, discover suppliers, manage clarification loops and prepare evaluation summaries while buyers keep the commercial judgement. In demand planning, AI could help capture qualitative market intelligence from sales, product and leadership teams without asking everyone to complete yet another spreadsheet. In PLM and capital approval, AI could help bring together structured data, expert comments, risk information and approval evidence so experts can focus on decisions rather than administration.
That, to me, is where the real opportunity sits. Not in pretending AI will instantly rebuild the whole supply chain. Not in applying it only to the most glamorous shop-floor examples. But in using it as a force multiplier across the office work that surrounds the supply chain.
AI as a force multiplier
I see AI as a genuine force multiplier in supply chains. We have already seen examples where AI can supplement procurement teams by gathering inbound delivery information, help improve process flows, enhance communication and support more effective supply chain interactions. There is also real potential in consensus demand planning, where AI can turn scattered opinions and conversations into a more structured, auditable signal.
This links directly to the AI-ready operating model work I have written about previously. The future is not simply about adding AI on top of existing processes. It is about redesigning work, so people focus on judgement, relationships, exceptions, creativity and decision-making, while agents support the administration, preparation, monitoring and response activity that too often slows everything down.
The danger is that businesses automate the wrong thing. If we simply make a poor process faster, we have not transformed anything. We have just created a quicker bad process. The better route is to start with the outcome, understand the friction, redesign the process and then decide where AI, automation and agents can help.
The real power of AI in the supply chain
For me, the real power of AI in supply chain is its ability to reduce administrative noise, improve communication, interpret and summarise complex information, and provide decision makers with better data at the point they need it.
That may sound less exciting than some of the grander claims made about AI, but it is far more useful. If AI can give buyers fresher supplier information, planners better market signals, customer service teams clearer answers, finance teams cleaner reconciliations and leaders earlier visibility of risk, then it will make a real difference.
AI transformation does not start with AI. It starts with understanding the work. Find the admin. Find the risk. Find the point where talented people are spending too much time winding the handle. Then use the technology where its talents match the job.
Because the supply chains that benefit most from AI will not be the ones that chase the loudest use case. They will be the ones that calmly identify where work gets stuck, redesign it properly and use AI to help good people do better work.
Want to learn more?
Book your free place at our next AI and supply chain workshop. A FREE workshop built for senior leaders who want to stop talking about AI in theory and start seeing where it delivers real gains across retail and supply chain.
