Your Data Knows What Happened. Can AI Help Predict What's Next?

That 69-point gap between belief and practice is where most forecasting efforts stall.

If you work in supply chain, procurement, finance, or operations, you already know how important cost forecasting is. You need to know what raw materials might cost next quarter. You need to plan supplier budgets. You need to estimate freight and logistics expenses. You need to understand how changes in demand could affect purchasing and inventory decisions.

And usually, you have plenty of data to work with. Historical costs. Supplier prices. Purchase orders. Freight rates. Demand forecasts. Currency data.

So why can forecasting still feel so difficult?

Why Forecasting Is Getting Harder

Because costs don't move in isolation anymore.

A forecast can look perfectly reasonable today and start looking very different a few weeks later. A supplier changes its pricing. Freight costs increase. A currency moves against you. Demand suddenly rises in one market. A geopolitical event disrupts a supply route.

Each change may seem manageable on its own. But when several happen together, their combined impact can be significant. And that creates a fundamental problem for today's supply chain leaders: how do you forecast costs when the variables driving those costs keep changing?

Now imagine trying to forecast that cost using only what happened last year. You might get reasonably close. But what happens when this year's conditions are completely different?

This is where traditional forecasting can become less useful. The question is no longer just: "What did this cost before?" It becomes: "What is changing around this cost, and what could those changes mean?"

The Problem Is Not Data Volume

For years, cost forecasting has relied heavily on historical data. And there is a good reason for that. Historical data tells you how costs behaved. It helps identify seasonality, supplier trends, demand patterns, and previous changes in spending.

But today's supply chains are dealing with a much more dynamic environment. Gartner describes geopolitical tensions, global trade and economic uncertainty as factors creating multiple potential outcomes for supply chain organizations. In a December 2024 survey of 506 supply chain leaders, Gartner found that only 19% of organizations fully integrated scenario planning into their supply chain strategies.

That tells us something important. Companies understand that they need to prepare for different possibilities, but many are still not fully equipped to do it. For a practical look at what readiness looks like before an organization invests in AI-driven forecasting, the six readiness pillars for retail and supply chain teams cover the data ownership, process standards, and decision-rights groundwork that makes any advanced tool actually deliver.

The challenge isn't simply having more data. It's being able to connect the right data and understand what it means for what comes next.

How AI Changes the Forecast

AI is not a crystal ball. And it shouldn't be positioned that way. The real opportunity is to use AI to process more signals, identify relationships between them, and continuously update the forecasting process as new information becomes available.

IBM describes AI forecasting as a process that can learn from historical data while incorporating new data as it arrives. It also highlights the ability of AI forecasting to work with more variables and identify complex relationships that may be difficult to capture through traditional approaches.

For a supply chain organization, that could mean bringing together:

Instead of looking at these datasets separately, AI can help identify the relationships between them. And that changes what a cost forecast can actually tell you.

The greatest friction point in scaling intelligent forecasting today is not the model itself, but the data foundation into which it is being deployed.

Everest Technologies, on supply chain AI adoption

From One Forecast to Many Scenarios

This is where the idea gets particularly interesting. Instead of producing a single number, you can explore what happens under different conditions:

Now you're not simply producing a number. You're exploring possible outcomes.

Gartner's 2025 research supports the growing importance of this approach: 88% of supply chain leaders surveyed agreed that scenario planning has a positive impact, while only 19% said it was fully integrated into their supply chain strategy. That gap is important. The value of scenario planning is increasingly recognized, but putting it into practice requires the right data, processes, technology, and skills.

The sequencing of those investments matters as much as the investments themselves. The best supply chain leaders settle the operating model question before any vendor RFP goes out - the same logic applies to data and AI strategy: define the model before selecting the platform.

AI Can't Fix Fragmented Data

This is probably one of the most important realities to understand. It is tempting to start with the AI model. But if supplier information is sitting in one system, freight data is somewhere else, market information comes from external sources, and important operational information still lives in spreadsheets, the AI model cannot simply make that fragmentation disappear.

The quality of the forecast depends heavily on the quality and accessibility of the data behind it. IBM specifically identifies data quality as one of the key considerations for AI forecasting. Its guidance also emphasizes that forecasting systems need ongoing monitoring because market conditions and data patterns change over time.

McKinsey makes a similar observation about supply chains, pointing to outdated infrastructure, fragmented data, and ongoing disruptions as barriers organizations need to address to get value from AI.

So a realistic AI forecasting strategy doesn't start with: "Where can we put AI?" It starts with: "Do we have the right data foundation for AI to actually help?"

Where Data Engineering and AI Work Together

This is also where the difference between an AI experiment and a useful business solution becomes clear. You need to first bring the relevant data together. Then you need to understand which variables actually influence cost. Then you can use AI and predictive models to identify patterns, test scenarios, and generate forward-looking forecasts.

The technology behind this can be sophisticated. But the business objective is simple: give decision-makers more visibility before the cost hits the business.

Where Everest Comes In

This is where Everest's expertise in Data Engineering and AI becomes particularly relevant. Everest helps organizations bring fragmented data together and create the data foundations required for advanced analytics and AI.

From there, AI can be applied to identify cost drivers, model different scenarios, and generate forward-looking insights. The objective isn't to replace the people making supply chain or procurement decisions. It is to give them better information to make those decisions.

For example, instead of a procurement leader seeing that projected costs have increased, the system could help surface the factors contributing to that change. Instead of waiting for a disruption to affect the forecast, teams can explore different scenarios and understand their potential impact. Instead of reviewing a static forecast every few weeks, organizations can move toward a more dynamic view that incorporates new information as it becomes available.

That is the difference between simply reporting costs and building cost intelligence.

The Goal Is Not Perfect Prediction

Let's be realistic. No AI model can predict every supplier decision, geopolitical event, currency movement, or sudden change in customer demand. The goal isn't to predict the future perfectly. The goal is to be better prepared for it.

A useful cost forecasting capability should help supply chain and procurement leaders:

• See important signals earlier
• Understand what is driving cost changes
• Explore different scenarios
• Make decisions before uncertainty becomes a problem

And that is where the conversation around AI in the supply chain becomes much more practical. It isn't about replacing the forecast that teams already use. It is about making that forecast more connected, more dynamic, and more useful for decision-making.

Turn uncertainty into cost intelligence.

Sources

1. Gartner, survey of 506 supply chain leaders on scenario planning integration, December 2024
2. Gartner, 2025 research on scenario planning impact in supply chain strategy
3. IBM, AI forecasting guidance: learning from historical data while incorporating new data as it arrives
4. McKinsey & Company, observations on fragmented data and outdated infrastructure as barriers to AI value in supply chains
5. Everest Technologies, Data Engineering and AI practice capabilities