From Insight to Action: How AI Is Changing Supply Chain Decision-Making

Traditionally, supply chain teams have used reports and dashboards to answer questions such as: What happened to sales? Where did inventory run short? Why were deliveries delayed?

Predictive analytics takes this a step further by using historical information and current conditions to estimate what might happen next.

Consider a retailer managing hundreds of products across multiple locations. Instead of relying only on past sales, its forecasting system could consider promotions, weather and other relevant factors to anticipate demand more accurately.

But a prediction alone does not create value. The value comes when the business uses that prediction to make a better decision about inventory, sourcing or delivery.

 

Not all AI performs the same role. Three categories help explain the difference.

• Predictive AI — What is likely to happen? It estimates future outcomes, such as demand, delivery times or inventory requirements.
• Generative AI — What can we create? It produces new content, such as supplier communications, summaries or draft recommendations.
• Agentic AI — What action can be taken? An AI agent can work through multiple steps in a process, use information from different sources and carry out permitted tasks towards a defined goal.

Imagine a procurement team reviewing supplier contracts.

Predictive AI might help estimate purchasing requirements.

Generative AI could draft a supplier negotiation letter.

An AI agent could analyze multiple contracts, identify price differences, prepare recommendations and draft communications for review.

The distinction is important: an AI agent does not simply provide an answer. It can help move a process forward.

The level of autonomy depends on how the system is designed and what authority the organization gives it.

Two examples shared by Sankararaman illustrate how AI can address real supply chain challenges.

Lowe's: Matching fulfilment decisions to customer needs

Appliances can be difficult to source and distribute because of supply constraints, long-distance transportation and complex fulfilment networks.

Sankararaman described how machine learning helped Lowe's distinguish between different customer needs. Someone whose refrigerator has suddenly stopped working may prioritise immediate availability. Another customer purchasing a premium appliance for a planned renovation may prefer a specific model and delivery date.

Treating both customers identically can lead to poor inventory and fulfilment decisions. Better predictions help the business respond to their different requirements without unnecessarily expanding inventory across locations.

The lesson: AI creates value when its insights reflect the actual business problem—not simply when it uses more data or more sophisticated models.

Novant Health: Finding opportunities in supplier contracts

Healthcare procurement can involve thousands of contracts, with different prices for similar products across suppliers and facilities.

At Novant Health, AI agents help analyse these variations, recommend potential price standardisation and draft supplier communications. Clinicians remain involved in validating recommendations, including considerations of instrument suitability and patient safety.

The lesson: AI can reduce time spent on repetitive analysis while allowing people to focus on decisions that require expertise and judgment.

These examples also reinforce an important principle: the simplest model that adequately solves a problem may be better than a more complex one.

One practical principle from the MIT CTL webinar is to distinguish between decisions that are routine, reversible and have bounded costs, and those that are strategic, difficult to reverse or high-stakes.

For example, an organisation might allow an AI agent to flag an inventory exception or adjust a replenishment order within pre-approved limits. A decision to replace a strategic supplier or enter a new market requires a different level of scrutiny.

Supply chain leaders can use three broad categories to guide their approach:

• Automate: Repetitive tasks with established rules and manageable risk.
• Delegate with controls: Routine decisions within approved limits, with monitoring and escalation when exceptions arise.
• Retain human authority: Strategic, high-impact or difficult-to-reverse decisions.

 

As AI takes on more routine analysis and execution, the role of supply chain professionals will evolve.

Sankararaman described this as a shift in work rather than the elimination of entire functions. As AI takes on more routine tasks, human judgment becomes increasingly important.

Four capabilities stand out:

• Supply chain fundamentals: Understand forecasting, inventory, procurement, logistics and network design.
• Business acumen: Connect operational decisions to service, cost, profitability and customer outcomes.
• Critical thinking: Validate AI outputs instead of accepting them blindly.
• AI literacy: Understand what the tools can do, ask useful questions and recognize their limitations.

You do not need to become an AI developer to benefit from AI. But you do need to understand the problem you are trying to solve and how to judge whether the answer is right.