Artificial Intelligence This is How AI Initiatives in Manufacturing Become Successful

Related Vendors

FAULHABER Drive Systems

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

Many executives today are ambitious about deploying AI comprehensively in manufacturing. However, without a robust data infrastructure, (pilot) projects will fail. Understanding that data should be treated as a fundamental infrastructure and not as a side issue is crucial for a fundamental transformation today.

In the industry, proofs of concept for AI initiatives are often very promising but then fail to deliver the expected return on investment. The cause typically does not lie in the AI models but rather in the data foundation: fragmented, poorly managed, and context-poor data environments, as commonly found in manufacturing.

Therefore, AI initiatives should start here with a radical realignment: data should no longer be understood as a byproduct of systems or as a compliance issue, but as the core of everything, from production planning to predictive maintenance. For this, data must be treated as its own infrastructure, maintained, standardized, and monitored for quality.

Gallery

Dozens of Different Data Streams

Especially in the manufacturing environment, this is no simple task. Dozens of different data streams are generated between industrial sensors, ERP systems, production lines, and supply chains. Merging these into a coherent picture that supports predictive or generative models is both a technical and cultural challenge.

The ability to contextualize is crucial here. For truly meaningful data products, companies need a holistic perspective. The foundation for this is consolidated datasets, consistent definitions, and clarity on who is responsible for which data.

Important: Company-Wide Definitions

The need for company-wide definitions is often overlooked but is crucial. For example, if different departments, systems, or applications use different definitions for "customer," this will lead to confusion and fail to provide reliable AI results.

The goal is not just to store more data but to create the conditions for its harmonization. This includes both technical tools and organizational maturity: governance, quality assurance, provenance tracking, access controls, encryption, and bias mitigation must be integrated into the data stack from the outset.

AI Agents Need Context

Currently, agent-based AI is associated with great promises. They can autonomously navigate digital ecosystems, query systems, and perform actions without human intervention. It is the next iteration of what was previously referred to as data federation or data virtualization. However, without a consolidated and well-managed data foundation that also provides the necessary context, there is a risk of lack of reliability and inconsistent responses.

In contrast to areas like finance or HR, manufacturing faces entirely different challenges due to the physical and distributed nature of industrial data: determining when a part needs to be produced, how to optimize uptime, or how to forecast availability requires deep integration with supply chains, operational processes, and equipment performance data. Without such a data structure, autonomous agents cannot function.

The Modern Industrial Data Stack

What should manufacturers prioritize when building their data infrastructure? Reference data, ERP systems, customer hierarchies, and supply chain relationships should come first. From here, they should quickly expand to IoT sensor streams and the ongoing collection, transmission, and analysis of production data in real time. At this point, data consolidation becomes essential.

Streaming platforms are becoming increasingly relevant. They make it possible to ingest data from various sources—be it machines, control systems, or external providers—and bring it into an analysis and AI environment almost in real time. It is crucial that these data streams are delivered in a common, open, and widely used storage format.

Data Lakes form the Foundation

High priority should be given to data lakes. They have significantly evolved in recent years and now form the foundation for reliable streaming, AI, and real-time decision-making. With robust technologies such as Apache Iceberg and Delta Lake, structured, scalable, and cost-efficient data access is made possible. Combined with unified catalogs and open

Tabular formats offer manufacturers the flexibility to run multiple workloads with a single clean and managed dataset. These advancements are gradually closing the gap between traditional data warehouses and data lakes. In the future, this distinction could even become completely obsolete.

Change in Technology And Corporate Culture

But even this is no guarantee of success for AI initiatives. The most challenging part is often not the architecture but the alignment: a traditional corporate culture—particularly in environments where IT, OT, and data teams work in silos—can doom even the most elegant solutions to failure. To change this, leadership must take the lead. An effective approach is to select a small but significant problem where data can make a measurable difference. If a positive result and a measurable ROI can be achieved here, it can be used as a case study to build internal trust and drive cultural change.

View Data As A Business-Critical Resource

As part of this cultural shift, it is also essential to address the question of responsibility for data. Data should not be viewed as a technical resource but as a business-critical asset. Governance mechanisms, ranging from data lineage and master data management to regulatory compliance and algorithmic transparency, form the strategic foundation for this.

Foundation for AI Success

Many manufacturing companies are still at the beginning of this journey. While some have already implemented modern data pipelines and data-centric strategies, others are still struggling with legacy systems, siloed teams, and limited pilot projects. However, the transformation is accelerating, with strong momentum around open-table formats and unified infrastructure in recent months. Companies should leverage this momentum, as technologies mature, to reassess their digital strategies and build something robust and scalable.

Ultimately, the winners in the field of manufacturing AI will not be those with the most elaborate AI models or the fanciest dashboards, but those who have taken the development of their data infrastructure seriously.

Tobias Knieper, Lead Marketing Manager, DACH bei Fivetran

Secure and Compliant Authentication in Laboratories

Secure & Contactless User Authentication at GMP/GLP