AI-Led Forecasting in the Supply Chain: Don’t Miss What’s Already Working<br><h3>By Daniel Stidsen</h3>

By Daniel Stidsen

Key takeaways:

• AI-led forecasting uses machine learning to automate model selection, incorporate real-time signals, and continuously refine accuracy.
• Organizations using e2open Demand Sensing have seen forecast accuracy improvements of up to 38%.
• Running AI and traditional forecasts in parallel is the most effective way to build trust and validate results before the full adoption.
• Machine learning-based forecasting is delivering measurable value now, and it’s one of the best places to start an AI journey in the supply chain.

Agentic AI continues to dominate the conversation in supply chain, and the excitement is well earned. Autonomous agents coordinating real-time decisions across complex networks, planners accessing insights through natural language, and machines executing multi-step workflows without human intervention all represent a compelling vision for the future.

Yet there is a gap between the vision and reality. Gartner's Balaji Abbabatulla, VP Analyst in the Supply Chain practice, predicts that by 2030, 60% of enterprises using Supply Chain Management (SCM) software will have adopted agentic AI features, up from just 5% today. But he is equally clear on the gap between availability and readiness. "While SCM tech providers will be delivering AI agents of various denominations to retain their competitive position in a rapidly evolving software market, supply chain data management, operations management, AI-readiness of the workforce, and network-centricity need to evolve to enable deployment of AI-led supply chain at scale," Abbabatulla said. In other words, the technology is coming, but most organizations are not ready for it yet.¹

That's exactly why where you start matters.

While many organizations are preparing their data, processes, and teams for the next generation of AI, one of the most valuable applications has quietly been delivering results for years: AI-led forecasting. Machine learning-based forecasting is already helping companies improve forecast accuracy, reduce inventory, and improve service levels. It remains one of the highest-impact, most accessible applications of AI in supply chain planning.

Four reasons to start with AI-led forecasting

AI-led forecasting is an excellent opportunity to embed AI into the supply chain. Not only because of the results it will create, but because of how it pressure-tests the culture, decision-making process, and adaptability to new technology.

1. AI-led forecasting can be validated in parallel with the existing process before it’s fully adopted

   Two forecasts can run in parallel: one using the existing, more manual process, and one using the forecast engine.  Then they can be compared directly. Neither will be perfect, but over time, the results will show which approach performs better. That's a binary outcome: either the existing forecast was more accurate, or the AI-led forecast was.

2. AI-led forecasting reveals where people are truly adding value

   Using AI to create a forecast means the decision is deferred to a machine, and that naturally raises questions about what the planner's role even is anymore. Traditionally, getting the forecast right meant going out to gather input, collaborating across the organization, applying judgment, and shaping the number. The machine works differently: it analyzes data directly and uses mathematical models to identify underlying patterns, rather than relying primarily on gathered inputs and assumptions. For the planner, the work doesn't disappear, but it does look much different. The role becomes focused on managing inputs to the forecast engine and monitoring the results. You'll gain clarity on where your effort is truly adding value, and where it might be unnecessary. 

3. AI-led forecasting is a stress test for how the organization actually makes decisions

   In theory, supply chain decisions should be driven by logic and data. In practice, it is human nature to lean on experience, comfort, and intuition. A person still decides when and how to use the AI forecast. But that decision reveals a lot. People want to trust the forecast. They need to be able to defend it. And when the machine produces a number that feels wrong, the instinct is to question it. Machine learning finds patterns people can't see, which means the "why" isn't always intuitive. That makes adoption feel risky, especially when anchoring to a single miss or outlier. But for many organizations, the greater risk is failing to adopt approaches that consistently outperform traditional methods.

4. AI-led forecasting is proven, and the numbers tell that story more clearly than anything else

   Time and time again, the data shows that AI-led forecasting provides consistent forecast accuracy improvements that deliver real value to the business. And that consistency is ultimately where trust gets built. Not through explainability or a clean story around the number, but through results that speak for themselves. Taking action on that can still be more difficult than expected, particularly in environments experiencing organizational change, evolving domain expertise, or unclear accountability. Recognizing these dynamics early, and addressing them alongside the technology, is often what separates successful AI adoption from stalled initiatives.

E2open Demand Sensing use case

AI-led forecasting is working successfully at many organizations using e2open’s long-term and short-term Demand Sensing solution. It’s a truly touchless forecast engine that improves forecast accuracy, strengthens service levels, and reduces the inventory buffers organizations carry for uncertainty.

A global CPG company manufactures and distributes products we use every day, including personal care, tissue, and healthcare items. They use e2open’s short-term Demand Sensing solution to achieve high forecast accuracy, which enables optimal distribution into the channel. 

The process works as follows: The company produces a demand plan by first creating a statistical forecast. The forecast is then enriched by the planner based on additional information: promotion plans, collaboration internally and externally, and more. When time advances into the near-term 7-week window, that demand plan is then sent to e2open for the Demand Sensing forecast to be created. The Demand Sensing forecast updates daily, incorporating signals including point of sale, store inventory, warehouse withdrawals, and open orders. It is completely touchless, operating based on the inputs it’s provided along with various parameters configured. That forecast then goes back to the execution system so it may be used to make distribution decisions, getting the right inventory to the right place at the right time. 

The results are significant. The Demand Sensing forecast improves the demand planning forecast by anywhere from 13%-38% depending on the horizon in focus. This means better product availability for customers when they go to pick up an item after a long day’s work. This means less buffer inventory required because the forecast is accurate upfront. And best of all, the process is largely touchless. Once configured, the engine automatically incorporates new signals and generates updated forecasts with minimal ongoing intervention.

A new generation of the e2open Demand Sensing solution

Those results were built on the previous generation of e2open’s Demand Sensing solution. Recently, e2open rebuilt the product from the ground up to support the growing size and complexity of global supply chains. The new architecture handles greater scale, gives planners more control and configurability, and enables continuous embedding of the latest AI advances as they emerge. The foundation got stronger. For customers already seeing up to 38% forecast improvements, the bar just got higher.

Want to learn more? Explore how e2open Demand Sensing can help your organization sharpen forecasts and support better execution across the supply chain.

The opportunity for AI in supply chains is already here

Agentic AI represents the next frontier. But while organizations prepare their infrastructure, data readiness, and culture to make it viable, machine learning is already delivering measurable improvement across supply chain planning environments. 

At e2open, we continue to invest across multiple advanced solvers: machine learning, optimization, heuristics, agentic AI and more; because supply chain problems don't all look the same. Agentic AI belongs in that mix. So does a touchless forecast engine that has delivered exceptional results for years. What matters most is having all the right tools available in our toolbox so we can match each solution to the problem it's designed to solve.

With e2open, our customers are well beyond asking, “Does this work?” The conversation has shifted to, “What do planners do when the machine is more accurate?” The answer is not simply nothing. It’s about optimizing how people and AI are working together to get the best outcome. More on that discussion another time.

¹ Balaji Abbabatulla, "Gartner Forecasts Supply Chain Management Software with Agentic AI Will Grow to $53 Billion in Spend by 2030," Gartner, April 7, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-04-07-gartner-forecasts-supply-chain-management-software-with-agentic-ai-will-grow-to-53-billion-in-spend-by-2030

FAQ: AI-led forecasting in the supply chain

What is AI-led forecasting and how does it work?

AI-led forecasting uses machine learning to predict future demand by identifying patterns in historical and real-time data. Rather than relying on manual input or static statistical models, the system evaluates multiple forecasting techniques — statistical and machine learning — and selects the best fit for your data. It continuously updates as new signals come in, including point-of-sale data, warehouse withdrawals, promotion plans, and open orders.

How is AI-led forecasting different from traditional demand forecasting?

Traditional forecasting typically involves spreadsheets, manual model selection, and input gathered from sales, marketing, and finance teams. It's time-intensive, difficult to scale, and introduces inconsistency through human judgment. AI-led forecasting automates model selection, incorporates a broader set of signals, and runs continuously, without requiring planners to manually shape numbers each cycle.

What data does AI-led forecasting need to be effective?

The more signal-rich the data, the better the results. Effective AI-led forecasting draws on internal data (shipment history, promotion plans, open orders) and external data (point-of-sale, store inventory, channel activity). Data quality and consistency matter more than volume. The system identifies which signals are most predictive and adjusts over time as those relationships shift.

What are the key capabilities to look for in an AI-led forecasting solution?

Look for automated model selection across statistical and machine learning techniques, the ability to incorporate both internal and external data signals, continuous model refinement across time horizons, configurability for edge cases, and minimal manual maintenance requirements. A touchless or near-touchless architecture is a strong indicator that the system is built for operational scale, not just proof-of-concept use.

What are the main business benefits of AI-led forecasting?

Forecast accuracy improvements are the most direct benefit — and they compound. Better forecasts reduce excess inventory, lower safety stock requirements, and improve service levels and product availability. For organizations using e2open's Demand Sensing, accuracy improvements of 13–38% have been documented depending on the planning horizon.

What are the risks or limitations of AI-led forecasting?

The most common challenges are organizational, not technical. When a machine-generated forecast conflicts with a planner's intuition, adoption stalls. Machine learning finds patterns that aren't always explainable, which can make it harder to build confidence quickly. Data readiness is also a factor, and inconsistent or low-quality inputs reduce accuracy. Addressing change management and data governance alongside the technology is essential for successful adoption.

How do you implement AI-led forecasting successfully?

Running AI-generated and existing forecasts in parallel is an effective way to build trust before full adoption. This creates a direct, auditable comparison over time. Success also depends on clearly redefining the planner's role, shifting from manual forecast shaping to managing inputs and monitoring outputs, and ensuring leadership alignment on what "trusting the machine" actually means in practice.

What is demand sensing, and how does it relate to AI-led forecasting?

Demand sensing is a short-term AI-led forecasting capability that updates daily using near-real-time signals like point-of-sale data, warehouse withdrawals, and open orders. It operates within a shorter planning window, typically days to a few weeks, and runs as a touchless process. It complements longer-range demand planning by sharpening accuracy precisely when decisions about inventory deployment and distribution need to be made.

Is AI-led forecasting the same as agentic AI?

No. AI-led forecasting uses machine learning to produce demand predictions — it automates a specific, well-defined task. Agentic AI refers to autonomous systems that can take multi-step actions, coordinate across workflows, and adapt in real time. Agentic AI builds on the foundation that AI-led forecasting helps establish, but requires a higher level of data readiness, infrastructure, and organizational maturity to deploy effectively.