Artificial intelligence is often presented as a tool for prediction, automation, and faster decision-making. In manufacturing, however, the success of industrial AI depends less on the sophistication of the algorithm and more on the quality of the data behind it. A conversation with Matthew Wright highlights a fundamental problem: many companies still lack the data infrastructure required to use AI effectively, which is why so many AI use cases in manufacturing never make it past the pilot stage.
The Data Gap at the Front End of Manufacturing
Much of the manufacturing industry’s technology investment has focused on the back end of operations, including machinery, production lines, factory equipment, and material-handling systems. The front end of manufacturing the stage where products are designed, ingredients or materials are selected, suppliers are identified, and formulas or specifications are established often still relies on outdated systems and processes.
In many companies, information about raw materials, product formulas, technical specifications, regulations, suppliers, and sales markets is scattered across multiple files and systems. The data may be incomplete, inconsistent, or difficult to access. As a result, even a simple question can be hard to answer:
“Which products contain this specific material?”
For example, suppose an ingredient such as nutmeg, honey, or a food coloring becomes scarce, restricted, or unavailable. A company should be able to immediately identify every affected product, find legal substitutes, and locate qualified suppliers. In many organizations, this process still requires manual searches, communication across several departments, and the review of large numbers of documents.
Why Industrial AI Struggles Without Reliable Data
AI can identify patterns, generate recommendations, and automate parts of the decision-making process. But it cannot consistently produce reliable answers from incomplete or disorganized information. If product and material data is inaccurate, the AI system’s output may also be inaccurate or misleading.
Before implementing advanced AI models, companies therefore need to create a structured and reliable foundational dataset. This dataset should clearly show:
- What materials and components each product contains
- Which suppliers provide each material
- In which countries the material is legally permitted
- What approved substitutes are available
- Which products would be affected by a material change
Once this foundation exists, AI can support more valuable applications, such as recommending substitute materials, predicting supply shortages, checking regulatory compliance, and reducing the time required to develop new products.
The Difference Between Market Analysis and Product Intelligence
Using AI to analyze consumer behavior is relatively straightforward. Companies can collect customer reviews, sales data, and changes in consumer preferences, then use AI to identify trends.
Understanding the internal structure of a product is much more difficult. For an AI system to recommend a substitute ingredient or material, it must understand product formulas, technical requirements, legal restrictions, costs, supplier availability, and market-specific rules. That level of intelligence is only possible when the system has access to accurate and well-structured industry data.
This creates an important distinction between “AI for analyzing consumer opinions” and “AI for operational decision-making.” The second application can create greater economic value, but it requires a much stronger data foundation.
The Economic Value of Structured Data for Industrial AI
Poor data management is not merely a technical issue. Delays in finding substitute materials, purchasing errors, production interruptions, and late product launches directly affect a company’s profitability. This is especially important in industries such as food and beverage, where profit margins are often narrow.
Companies sometimes hide these inefficiencies by adding more employees to the process. More people are assigned to search for information, coordinate between departments, and manage changes. However, this approach does not solve the root problem: the organization’s data remains fragmented and difficult to use.
Technology creates real value when it shortens complex processes, accelerates decisions, and reduces the risk of human error. Under these conditions, AI can become a tool for improving profit margins rather than simply an impressive or fashionable technology.
Conclusion: Data First for AI in Manufacturing
The most important lesson for manufacturers and startups working in this field is that successful AI adoption does not begin with building a model. It begins with understanding, collecting, cleaning, and structuring data.
AI can accelerate product development, improve supply chain management, and reduce costs only when it is built on accurate, accessible, and trustworthy information. Without that foundation, even the most advanced algorithms will struggle to solve real operational problems.
The future of industrial AI in manufacturing will belong to companies that first invest in strong data infrastructure and then apply AI to important business problems with measurable economic impact.
