Artificial intelligence offers JD Edwards users opportunities to improve forecasting, streamline operations, and make better business decisions. But turning those opportunities into practical solutions requires more than selecting an AI platform. It requires reliable data, well-defined processes, and a clear understanding of where automation delivers value.

At INFOCUS 2026, Tom Messer, Senior Director of ERP and Manufacturing Systems at Mueller Water Products, shared how his organization is approaching AI adoption across its JD Edwards environment. Joined by Mo Mohammed Shajat, VP of Applications and Advisory Services at ERP Suites, Messer outlined Mueller’s strategy for applying AI to demand planning, production scheduling, and operational decision-making while keeping employees in control.

Their message was straightforward: Start with meaningful business problems, improve the processes behind them, and use AI to help people work more effectively.

Building an AI Strategy Around JD Edwards

Mueller Water Products manufactures fire hydrants, valves, and related infrastructure products, operating foundries and distribution facilities across North America. Its manufacturing environment involves complex inventory requirements, production constraints, and coordination among sales, operations, and supply chain teams.

Messer described three approaches to AI adoption within Mueller’s technology environment. The first involves tools such as Microsoft Copilot, which Mueller uses for document retrieval and ServiceNow-related assistance. The second focuses on standalone AI applications connected to enterprise data, including the company’s demand planning initiative in Amazon Quick. The third, and Messer’s longer-term priority, involves embedding AI capabilities directly into JD Edwards.

Rather than forcing users into separate applications, Messer wants AI-generated insights and recommendations available through familiar JD Edwards interfaces, including CafeOne pages and widgets.

Mueller is following an Inform, Assist, Act progression. Initially, AI provides information and identifies exceptions. Next, it recommends corrective actions. Eventually, users could approve suggested changes that the system executes automatically. For now, Mueller is emphasizing human oversight rather than autonomous transactions.

The supporting architecture combines JD Edwards, an AWS Redshift data lake, orchestration, and secure API middleware. Authentication and encryption provide a controlled gateway between enterprise systems and AI services, allowing Mueller to evaluate different models without committing its entire strategy to one vendor.

Replacing Spreadsheet-Heavy Demand Planning With AI

Mueller’s most advanced initiative focuses on demand forecasting and sales, inventory, and operations planning (SIOP). At the time of the presentation, the solution was approaching prototype user acceptance testing.

Previously, planners relied on extensive spreadsheets to consolidate historical sales, open orders, Salesforce opportunities, inventory, and other information. One particularly complex workbook contained approximately 75 tabs.

The problem wasn’t that planners lacked expertise. It was that assembling information consumed time they could otherwise spend evaluating business conditions.

Messer explained that planners could perform many sophisticated analyses manually, but rarely had enough time to complete them all. AI offers an opportunity to accelerate those calculations and make more scenarios available.

Built primarily in Amazon Quick and connected to Mueller’s data lake, the solution uses AI agents and statistical forecasting models to evaluate historical demand alongside external indicators such as housing starts, weather, and interest rates.

Importantly, Mueller isn’t expecting AI to deliver a single unquestionable forecast. Planners can compare historical trends, statistically generated forecasts, and forecasts adjusted for external economic conditions before selecting or refining the final plan.

Early prototype feedback reinforced the importance of transparency. Planners wanted to understand why a forecast changed, not simply see a revised number. Consequently, Mueller is separating forecasting layers so users can trace adjustments and apply their own experience.

The solution also supports scenario planning around inventory and customer service targets. For example, planners can evaluate how increasing a service-level target might affect required safety stock and inventory investment.

Once finance, operations, and supply chain agree on a feasible plan, the approved forecast can be exported into JD Edwards to support master production scheduling and material requirements planning (MRP).

Bringing AI Into Production Scheduling

Mueller’s next major opportunity is production planning, where the company wants to combine JD Edwards transactions with manufacturing execution system (MES) information.

The envisioned solution would incorporate inventory, work orders, machine availability, cycle times, overall equipment effectiveness (OEE), and production capacity. AI could then recommend scheduling adjustments based on equipment constraints and operational priorities.

This addresses an important limitation of traditional JD Edwards planning: MRP operates with infinite-capacity assumptions, identifying requirements without necessarily resolving production capacity conflicts.

Messer wants AI to help answer practical questions. If a machine is overloaded, could another machine handle the work? Should an order be rescheduled? Which production sequence minimizes changeovers?

Accurate operational data remains essential. Messer recalled visiting Mueller’s Chattanooga facility and watching a planner move from the JD Edwards shop floor workbench into a spreadsheet, then manually verify inventory because system quantities weren’t trusted.

That experience highlighted a foundational challenge: AI cannot reliably optimize production when inventory balances, work order statuses, and manufacturing transactions are inaccurate.

Mueller is working to improve transaction accuracy and integrate MES information before expanding AI-driven scheduling.

The longer-term goal extends beyond internal efficiency. More reliable production planning could support capable-to-promise functionality, helping customer service representatives determine when an out-of-stock product can realistically be manufactured and delivered.

Approved schedules could also flow into MES, with planners retaining authority over changes. Eventually, AI-assisted workflows might update work orders or prepare supplier purchase orders, subject to appropriate approval controls.

Choosing AI Use Cases That Deliver Business Value

Shajat emphasized that organizations should examine entire business processes before introducing AI. Automating an inefficient process without reconsidering its steps may simply preserve unnecessary work.

His approach prioritizes processes that occur frequently, involve repetitive information gathering, contribute to valuable business outcomes, and require human judgment.

For example, running reports, comparing inventory screens, and gathering information are necessary activities, but they don’t necessarily create the same value as deciding how to fulfill a critical customer order.

AI can reduce the effort required to gather and interpret information, allowing employees to concentrate on decisions.

Shajat also stressed the importance of capturing institutional knowledge. Experienced employees often understand business exceptions and unwritten rules that aren’t documented anywhere. Recording their workflows and decision-making logic can help organizations translate that expertise into reusable AI-assisted processes.

Employee involvement is equally important. Teams are more likely to support AI initiatives when they understand that the objective is to eliminate repetitive tasks and strengthen their capabilities rather than disregard their expertise.

Data Quality, Cost, and the Importance of Starting Small

The discussion also addressed two persistent AI adoption challenges: unreliable data and uncertain operating costs.

Messer acknowledged that Mueller hadn’t yet established actual production costs for its demand planning solution. Usage would need to be monitored after deployment.

Shajat explained that thoughtful architecture can help control expenses. Not every calculation requires a large language model. Statistical modeling and conventional machine learning can perform certain forecasting tasks more economically, while AI agents handle work requiring additional context or reasoning.

Data preparation may prove even more demanding. Inconsistent product hierarchies, incomplete inventory transactions, and undocumented processes can undermine automation regardless of the technology selected.

However, neither presenter recommended waiting for perfect data across the enterprise. Instead, organizations should identify manageable opportunities where business teams are motivated, information is reasonably reliable, and results can demonstrate value.

What JD Edwards Organizations Can Learn From Mueller

Mueller’s experience demonstrates that successful AI adoption is less about introducing sophisticated technology than improving how employees access information and make decisions.

Its demand planning initiative is progressing toward testing, while production scheduling remains a developing vision. Both illustrate a practical approach: preserve JD Edwards as the operational foundation, use AI to accelerate analysis, and introduce automation gradually with appropriate controls.

For organizations evaluating AI with JD Edwards, the most useful starting point may be a familiar frustration: a spreadsheet-intensive process, an inventory discrepancy, or a recurring decision that requires gathering information from multiple systems.

As Messer emphasized, organizations can spend endless time evaluating platforms and possibilities. Meaningful progress begins when they select a practical use case, build something, and learn from the results.

Want more?

Explore more content and resources to help you get the most from your JD Edwards investment:

• Visit the Quest Learn Library for blogs, how-to demos, and on-demand sessions.
• Connect with peers in one of our Quest JD Edwards Community User Groups to swap stories, ask questions, and share tips with other users facing the same challenges.
• Bookmark LearnJDE.com for updates from JD Edwards development.

Not a Quest member yet? Join today and tap into the ultimate Oracle customer network.