Your supply chain now faces more variables than just a few years ago. The challenge is no longer identifying the volatility, but bridging the gap between sensing that something is wrong and fully understanding what changed, what it means, and how you plan to respond while there’s still an opportunity to act.
Artificial Intelligence (AI) fits into that gap as a practical tool, not a side experiment. It helps teams spot patterns earlier, connect cause and effect, and choose a course of action with more confidence in the middle of everyday work.
Here, we take a fresh look at that potential, and if you want to go deeper, let us point you to our new eBook, Using AI to Meet the Challenges of the Modern Supply Chain.
AI in the Supply Chain Feels Both Urgent and Unfinished
If you talk to peers, you probably hear two things at once: Everyone is experimenting with AI, and very few feel finished.
Recent research backs that up. One recent survey found that 95% of manufacturers already use AI for some aspect of supply chain management, often for efficiency and planning. Yet more than 90% also say legacy technology still gets in the way of their AI ambitions.
Gartner tells a similar story from another angle. Only 23% of supply chain organizations report having a formal AI strategy, even though leaders see clear potential. Short-term ROI pressure often wins out over longer-term transformation.
So, organizations often end up with pilots that prove a concept – a better forecast here, a smarter alert there – but do not yet change how the broader network runs. The question becomes: where do you focus next so AI stops being a side project and starts improving how you move, store, and promise product?
A Real-world Example: Forecasting Ice Cream with the Weather
To answer that, it helps to look at a concrete example.
Unilever’s ice cream business faces a classic demand problem: sales spike and drop with the weather. In many markets, a cool weekend can leave freezers full. A surprise heat wave can lead to empty shelves and unhappy customers.
Unilever teams began using AI models that blend internal data with external signals such as local weather forecasts and events. In Sweden, for example, they feed in temperature data and store-level sales to fine-tune volume forecasts for brands like Magnum and Ben & Jerry’s.
The result is more accurate forecasts and a different day-to-day reality. Stores receive the right mix of products ahead of warm days, improving on-shelf availability and reducing waste when temperatures drop again. Planners, sales teams, and logistics partners share a more realistic view of demand and can adjust transport, production, and promotions with more confidence.
You may not ship ice cream, but you face your own version of this challenge – a set of variables traditional tools handle poorly, where better pattern-finding and faster feedback would help.
Three Conversations to Start with Your Team
Rather than asking, “What can we do with AI?” it often helps to ask where better signals and faster interpretation would change day-to-day decisions.
Here are three conversations that tend to open up productive ideas.
- How good is our view of demand, really?
Traditional demand forecasts lean heavily on historical averages and a few known seasonal patterns. That approach struggles when markets fragment, lead times fluctuate, or promotions move quickly across channels.
In one recent study, AI-based forecasting models reduced errors by up to 50% and lowered logistics costs by roughly a third. Improvements of that magnitude translate into fewer lost sales, less excess inventory, lower warehousing costs, and leaner administrative effort. - Which risks do we want to see earlier?
Risk shows up in small ways long before it becomes a headline. A carrier begins missing pickups on a lane. A supplier stretches lead times. A region introduces regulations that add friction to inbound flows.Many manufacturers see AI and related tools as a way to manage that complexity. AI can continuously scan data for patterns that match emerging risks – changes in lead times, defect rates, transit times, or order patterns – and tie those patterns back to specific customers, products, or financial exposure. Instead of discovering a problem when a customer escalates, your team sees a clear signal with context and options to discuss. - Where are people chasing data instead of solving problems?
Supply chain teams spend a surprising amount of time gathering data – exporting reports, reconciling figures from different systems, and preparing slide decks. That work matters, but it often leaves little energy for focused problem-solving.Here, AI’s role is less about prediction and more about interpretation. In Microsoft Dynamics 365 Supply Chain Management, for example, Copilot features can summarize long lists of work orders, purchase orders, or shipment records, highlighting the items that need attention and explaining why.
A practical way to move forward
AI provides the most value when it processes more signals than people can easily monitor, converts those signals into clear and specific options, and operates close to the work inside the tools your teams already use.
To get there, you do not need a massive program on day one, but you do need to begin taking deliberate steps.
Keep Exploring: AI and the Modern Supply Chain
To help you sort through options in a Microsoft environment, we invite you to check out our eBook, Using AI to Meet the Challenges of the Modern Supply Chain. The guide walks through concrete scenarios, shows how AI capabilities surface inside Dynamics 365 Supply Chain Management, and outlines a path that respects both your current realities and your long-term goals.
You can use it as a conversation starter with your own team – and, if it is helpful, as a starting point for a deeper discussion with Velosio about where AI can do the most good in your supply chain.