Manufacturers are under increasing pressure to “do something with AI.” Vendors promise intelligent systems, boards ask about innovation, and competitors seem to be moving fast. But for middle-market manufacturers, the real question isn’t whether artificial intelligence (AI) matters — it’s how to apply it in a way that creates real, durable value. The answer starts with a clear understanding of how AI works alongside your enterprise resource planning system (ERP), and why strong data foundations — not technology widgets — ultimately determine success.
What do we mean by AI in the manufacturing context?
When manufacturers talk about AI, they’re often lumping very different capabilities into a single label. In reality, much of the value manufacturers are seeking comes from advanced analytics and predictive tools that have existed for years in areas like forecasting, maintenance, and planning — using ERP data to improve decisions. These tools are becoming more powerful as organizations standardize and integrate larger volumes of data.
More recently, generative AI has entered the conversation, emerging around ERP systems to help team members find information, interact with reports, and access documentation more efficiently. In both cases, AI works best not as a standalone solution, but as an extension of the ERP, amplifying insights and automation rather than replacing core systems.
Common ERP driven use cases include:
- Demand forecasting and sales planning.
- Supply chain and inventory optimization.
- Predictive and preventive maintenance.
- Financial and operational insight automation.
- Workflow automation built on ERP structures.
As data maturity increases, more advanced use cases can be considered, such as margin analysis, identifying unprofitable orders, or flagging operational patterns that deserve management attention. The key is not pursuing AI for its own sake but tying it directly to real business problems that already exist.
The role of the ERP in AI strategy
The ERP is the backbone of a manufacturer’s AI strategy. It defines how data is structured, captured, and governed, serving as the system of record for materials, bills of material, routings, labor, costs, customers, suppliers, and financials. In effect, the ERP acts as the template for what AI will see and interpret.
Why do many AI initiatives fail? The biggest gap to AI readiness isn’t the ERP software itself — it’s the quality and consistency of the data held within it. A common mistake is attempting to deploy AI without first doing the hard work of data cleanup and governance. There’s an old saying in data analytics: “garbage in, garbage out.” With AI, the stakes are even higher. Poor or ungoverned data doesn’t just lead to bad answers — it leads to “garbage in, poison out.” AI will confidently deliver fast responses that appear credible but are inconsistent, misleading, or difficult to trust.
This problem often shows up when data is poorly defined. For example, asking AI, “Which location had the top sales last month?” may return a correct answer until the same question applied to a different period quietly switches to a different definition of “sales.” To a human, nuances like these can often be inferred. To AI, they must be explicitly defined. The output may look fine from the user’s perspective, but, without governance, the results are no longer consistent or auditable. As your company moves forward with AI, these nuances stop being minor details and become essential to producing results that are trustworthy, governed, and repeatable.
Building a strong data foundation
Many middle-market manufacturers are building an integrated data environment for the first time and are still untangling spreadsheets, legacy systems, and inconsistent definitions. AI is not an easy button that shortcuts this work. What makes a strong data foundation for AI?
A strong data foundation must include:
- Well-governed master data (e.g., items, customers, vendors, BOMs, routings).
- ERP configuration aligned with actual business processes.
- Integrated systems with a trusted source of truth.
- Clear data ownership, stewardship, and accountability.
- Defined metadata and documentation that explain how data should be interpreted.
To learn more about building a strong data foundation, read our article “Foundational data considerations for generative AI.”
How to get started
A practical path forward starts with realism. Begin by identifying compelling business problems — not AI features — to address. Look for existing operational or financial pain points that are tangible enough to motivate your teams and justify the effort required to improve data quality.
Next, assess whether your data can actually support the selected use case. If it can’t, the priority should be governance, structure, security, and process control inside the ERP, not AI experimentation.
Then, pinpoint an area with relatively strong data and pursue a quick, achievable win. Preventive maintenance is often a strong candidate because the required data set tends to be smaller, more mature, and already well established in the ERP. Quality trends, scrap analysis, and focused supply chain forecasting are also common starting points.
Crucially, define what “success” means upfront. A quick win isn’t about simply turning something on — it’s about solving a meaningful problem with measurable return on investment. Early success builds momentum and reinforces that strong data foundations are what make progress sustainable.
Finally, expand thoughtfully as your data maturity increases. Higher impact initiatives such as demand forecasting, logistics optimization, workflow automation, and advanced financial insights can follow, sequenced deliberately rather than pursued all at once.
Not all AI is enterprise-ready
The AI market is flooded with niche tools and bold promises. Many vendors are injecting AI into specific applications that work well in isolation but can become unwieldy when broader data is introduced. Be cautious of:
- One-off AI tools that create an accumulation of shortcuts, workarounds, or outdated technology that can make your systems harder, slower, or more expensive to change over time.
- Solutions that operate outside of your ERP governance.
- Products marketed as “AI” that are little more than wrappers around tools you may already own.
- Claims that ignore your current position on the data maturity curve.
The reality is that while ERP vendors have used predictive analytics for years, AI-ready ERP systems are not fully here yet, particularly when it comes to embedded generative AI. Despite impressive demos and roadmaps, most organizations aren’t yet using conversational AI to transact directly inside their ERP day to day. Instead, generative AI is beginning to emerge around ERP systems — supporting reporting, knowledge access, training, and analysis — rather than replacing core transactional processes. That distinction matters when setting expectations, prioritizing investments, and separating near-term reality from longer-term vision.
Should I hire an advisor? Practical guidance for a disciplined AI journey
An experienced management consultant can help you through AI hype and focus on practical value. Advisors with deep manufacturing experience bring an objective perspective, assessing ERP configuration, data governance, security, and process controls to help sequence AI initiatives based on real readiness. Equally important, they’ll help your team avoid rushed implementations that stall, fail, or erode trust in data while still enabling visible, credible progress. If you’re feeling pressure to innovate, the right guidance can turn intention into results without compromising the foundations.
AI doesn’t change the rules — it follows them
AI can be transformational for manufacturers but only when it’s built on strong ERP systems and governed, reliable data. It accelerates the work you’re already doing well, but it doesn’t fix broken foundations. Companies that see real value from AI first invest in data maturity, analytics, and accountability. The most successful efforts start small: one part of the system, one business problem, and a clear measure of success then expand deliberately. Manufacturing is built on precision, discipline, and reliability. AI succeeds the same way — not through shortcuts but by doing the fundamentals well.
