According to the industrial software provider IFS, many AI initiatives in manufacturing fail due to typical challenges in operational practice. The company identifies five common barriers that hinder the productive use of AI.

Poor data quality

According to IFS, the biggest weakness in AI projects is usually the quality of the data. Fragmented data landscapes—consisting of a mix of data that has accumulated over time from a wide variety of sources—are typical. This data is often incomplete, unsynchronized, or interpreted in different ways. Therefore, AI projects should initially rely on a core set of selected operational data—such as data from maintenance or quality assurance—which can then be optimized step by step.

Fears and Latent Rejection

AI projects are often met with reservations. The reasons for this include opaque decision-making processes, the fear of losing control, or uncertainty about how the technology will affect one’s own role. According to IFS, these factors are often underestimated at the management level, and necessary change processes are neglected. Therefore, part of the effort to win people over involves making project successes transparent. Acceptance arises when employees recognize the concrete benefits of the technology and see that it delivers helpful results.

Lack of Objectives and Use Cases

According to IFS, many AI business cases are created solely for the purpose of making an investment decision. However, once a project begins, they are often neither actively monitored nor verified against actual operational data. In practice, there is often a lack of clearly defined key performance indicators, as well as a lack of continuous comparison between planned and actual benefits. As a result, the expected success often remains an assumption rather than a measurable metric.

Lack of practical applicability

Another obstacle is the transferability of pilot projects to day-to-day operations. Many projects are designed as “lighthouse” initiatives without being integrated into existing systems and decision-making processes. As a result, they work in the lab but not in actual operations. Causes include, among other things, dependence on specialists, insufficient user-friendliness, or limited use by early adopters.

Lack of Governance

IFS identifies a lack of governance as one of the most common barriers to scaling. Many companies launch AI initiatives on a project-by-project basis without establishing clear responsibilities for data quality, model operation, business decisions, or issues of liability and compliance. There is a lack of clear responsibilities in both IT and business units, a lack of collaboration in cross-functional teams, and a lack of a defined division of labor between humans and AI.

“Successful AI projects are characterized by three key features,” explains Sören Michl, Vice President of AI Adoption at IFS. “First, they are designed from the outset not as experiments but as systems that boost productivity; second, they rethink processes to make the best possible use of AI; and third, they integrate AI directly into existing workflows.”