QAD | Redzone (QAD) used its Champions of Manufacturing event in Chicago to introduce a series of AI capabilities, including role-based agents it calls Champions. The point is not how many agents QAD announced. It is how AI, ERP, plant operations, supply chain and workforce systems can share context and support decisions across the business.

At the center is Manufacturing Intelligence, which QAD describes as the industry’s first federated intelligence layer for manufacturing. Federated intelligence does not require everything to live in one system or move into one platform first. The goal is to connect context across ERP, the factory floor, supply chain, quality and workforce systems, including technology outside QAD.

This is a familiar problem for manufacturers. A machine issue can become a quality problem, affect a production order, change inventory availability and delay a customer shipment. The information needed to understand that chain is often split across ERP, manufacturing execution, quality, asset and supplier systems, along with the people closest to the operation.

The industry is already spending to address this problem. Deloitte’s 2026 Manufacturing Industry Outlook found that 80% of surveyed manufacturing executives plan to put at least 20% of their improvement budgets into smart manufacturing. The harder part is turning fragmented data and operating knowledge into better decisions.

That is the real test for manufacturing AI. Can it connect enough trusted context to understand what is happening, what is at risk and what should happen next? I explored that foundation further in InfoWorld in Why Trusted Context Is Becoming The Currency For Enterprise AI . Disclosure: KramerERP provides paid research, advisory and consulting services to technology companies, including ERP and data vendors listed in this article.

NVIDIA Brings AI Closer To Manufacturing

QAD’s approach combines NVIDIA’s computing, vision, multimodal AI, optimization and edge capabilities with its manufacturing context across ERP, frontline operations and thousands of plants. Initial applications include visual inspection, document processing, conversational access to manufacturing information and production optimization.

Visual inspection is an easy way to see how this could work. A camera can spot an anomaly, but the alert alone is incomplete. The value comes when the system connects what it sees to the production order, product, lot, supplier, asset and quality process, then helps the team understand the impact while there is still time to respond.

Manufacturing also changes the AI infrastructure math. Some workloads need low latency and local processing. Others can use larger models and more reasoning. QAD is evaluating NVIDIA NeMo Switchyard to route routine workloads to more efficient local models while sending more complex reasoning to larger models across cloud, data center and edge environments.

The model should fit the work manufacturers are trying to do. A plant floor decision may need speed and local processing, while a more complex decision may require more reasoning. Cost, uptime, governance and data sovereignty should all be part of that decision.

Manufacturing AI Moves Into Daily Work

QAD also used Chicago to show how AI is moving into day-to-day manufacturing work. Champion Assist brings natural language interaction into Adaptive ERP, while other Champions support areas such as procurement, sourcing and accounts payable. Lynx Champion, QAD’s AI-based ERP modernization tool, addresses a different problem by analyzing legacy customizations and helping manufacturers decide what should stay, what can be retired and what needs to be modernized..

Redzone brings AI closer to the plant floor, where issues such as changeovers, downtime, quality and training are happening in real time. The product names are less important than the connection between the systems. ERP provides the order, inventory and financial context, while frontline systems add what is happening in production right now.

That is likely where many manufacturers will start, not by automating an entire plant at once, but by improving frequent decisions around production, quality, maintenance, materials and labor.

QAD is also expanding the connection between Redzone Connected Workforce and Adaptive ERP. Built with Boomi, the integration can connect ERP, enterprise asset management, manufacturing execution and quality systems, creating a two-way flow between enterprise planning and the plant floor.

This is where the move from system of record to system of action starts to take shape. ERP still records the transaction, but AI can use that business context together with current operating conditions to help people decide what to do and move the work forward. AI value does not arrive alongside ERP. It arrives through it. I explored that evolution in more detail in Forbes in Why ERP Became The Execution Layer, Not Just The System Of Record .

Industry 5.0 Keeps People In Control

The Trade Compliance Champion, QAD’s AI tool for trade classification and compliance workflows, shows how work can be divided between AI and people. Product Classification can prepare Harmonized Tariff Schedule (HTS), Export Control Classification Number (ECCN) and other export classification recommendations with confidence scores and rationale. The AI prepares the recommendation and audit trail, while a trade professional makes the final determination.

That is also how I think about Industry 5.0. It is not another technology label. It is an operating model for how people, automation and AI divide work, authority and accountability. Technology can take on repeatable work and preserve operating knowledge, but judgment and accountability still need an owner. I explored that broader operating model in Forbes in Why AI Requires A New Enterprise Operating Model .

The workforce side is hard to separate from the technology story. Deloitte and the Manufacturing Institute recently highlighted technician shortages and AI’s potential to embed more knowledge into day-to-day work. Deloitte also estimates that more than 81% of manufacturing task hours will remain human driven.

Technology enables transformation. People determine the outcome. The goal should not be to remove people from every decision. It should be to use their judgment where it adds the most value while technology handles more of the routine work.

Manufacturing AI Competition Moves Toward Execution

QAD is not operating in a vacuum. Epicor Prism is bringing vertical AI agents directly into ERP workflows. Infor now has more than 100 Industry AI Agents along with its Agentic Orchestrator. IFS Loops is taking a Digital Worker approach to industrial operations. SAP is embedding Joule Assistants across planning, manufacturing, logistics and asset management. Oracle is bringing AI agents and agentic applications into Fusion Cloud Supply Chain and Manufacturing.

Microsoft’s manufacturing strategy spans Dynamics 365, Copilot, Fabric and Azure, with agentic ERP use cases across demand, supply, production, fulfillment and field service. The common idea is to move business context closer to operational decisions.

The competition also extends closer to the plant floor. Siemens is working with NVIDIA on industrial AI and AI-driven adaptive manufacturing. Rockwell Automation is extending Plex MES with FactoryTalk ResilientEdge, bringing local edge execution together with cloud intelligence and orchestration. Different vendors are coming at this from different directions, but the question is similar: How do you connect enterprise and plant context, then turn insight into governed action?

The question is not who has the most agents. It is who can make all of this work across manufacturing environments and show the outcome. QAD’s federated approach fits the reality that manufacturers run mixed systems, but that also makes interoperability a tougher test. It has to work across manufacturing execution, product lifecycle management, asset, automation and non-QAD ERP systems or it risks becoming one more integration layer to manage.

Federated Intelligence Connects Context To Execution

The Chicago announcements are easier to understand as one execution architecture than as a list of AI features.

Adaptive ERP provides the transaction and process foundation. Redzone adds what is happening on the plant floor. Supply chain and trade applications bring in more business context. NVIDIA adds visual AI, computing and optimization, while ChampionAI can use that context inside workflows. Manufacturing Intelligence is intended to connect it all.

This is why I continue to see ERP moving toward an enterprise execution layer. The system of record remains foundational because it holds the transactions, processes and business history the enterprise depends on. The value comes when that context supports a decision and moves it into action. That is the progression from system of record to system of action and now toward a system of intelligence.

Manufacturing AI Must Prove The Outcome

Architecture is easier to explain than execution. QAD expects Manufacturing Intelligence to reach the market in spring 2027, so much of the vision still needs to move into production. The company plans structured customer deployments tied to operational and financial benchmarks with a goal of measurable value within 90 days.

Three things will tell me whether this works. Can the model repeat across plants and industries? Can it operate across mixed technology environments without adding more complexity? And can agents take on more work while keeping human oversight, auditability and recovery in place?

Manufacturers already got the memo. Throughput, downtime, quality, inventory, working capital, productivity, service, operating cost and margin tell us more than the number of agents deployed.

System of record → trusted context → decision → action → measurable business outcome.

If QAD can shorten the path from context to action and show repeatable results across manufacturing environments, the federated intelligence layer becomes more than another AI architecture. Start with a process where performance can be measured and ownership is defined. Show the result, then scale what works.