The top 5 reasons why AI projects fail in production

In fact, there are a number of essential aspects that need to be considered when integrating AI. The following five reasons show why AI projects in production can fail and what can be done about it.

The integration of AI into industrial applications makes it possible to optimize processes, improve product quality and reduce operating costs. AI can support well-founded decisions and react proactively to changes, for example in predictive maintenance, quality control, production planning and control as well as robot automation.

The effort required to integrate AI depends on the complexity of the production processes, the degree of digitalization, the specific requirements and the existing infrastructure. Several steps are necessary, from planning and employee training to technical implementation and system optimization. To ensure that everything runs as smoothly as possible, the following five aspects should be avoided and corrected as follows.

Reason 1: Insufficient data quality and structure

One of the biggest hurdles for many industrial companies is inadequate data quality. This is because the quality of the results generated with AI stands and falls with this. Data is often inconsistent, incomplete or available in unstructured formats.

However, experience has shown that up to 80 percent of the effort required for a viable AI data product lies in data preparation. Fortunately, this can be professionalized and, above all, automated. Robust data pipelines that are created using data engineering are important. Data analytics and data science are then crucial. Data analytics is used to provide and analyze data automatically. This provides significant support in the decision-making process. In the next step, data science recognizes recurring patterns, enables predictions and can use the data to solve complex problems. This usually does not work with the sole use of an intelligent tool, but requires an end-to-end data strategy.

Reason 2: Lack of understanding and expertise in technology

Another obstacle is a lack of understanding of the technology and a lack of in-house expertise. If companies nevertheless dare to implement AI themselves, there is a good chance that the project will fail. This is because the market for AI tools is dynamic and confusing, which is a challenge even for experts. Working with an AI service provider can help. Large companies in particular often fear that external service providers will not be able to grasp their complex production processes straight away or generally have reservations about disclosing their own expertise. In this case, it is advisable to first define an AI use case and work with an expert in a workshop. This allows companies to gain an impression of the expert's skills and establish initial AI basics and knowledge for themselves.

At the same time, IT managers should start acquiring AI knowledge at events, trade fairs and via webinars. It is also always possible to participate in publicly funded AI research projects. The first AI use case is then a small, agile start. Companies should actively involve everyone involved from the outset and provide training and build up AI expertise across all project phases.

Reason 3: Complex production environments, outdated structures

In complex production environments with outdated structures, the implementation of digitalization and innovation projects is a particular challenge that can block many things right from the start. A good starting point here is the use of proven analysis, assessment and evaluation processes. For example, the complexity index or value stream management identify various optimization potentials. Step-by-step digitalization in the agile process, which involves all employees, then has a promising basis for success. It is essential to think specifically about what the AI use case should achieve and how this performance can be made measurable. External support can also be useful here - ideally not just for the implementation of solutions, but for holistic support in the transformation process. The good thing is that if companies approach such a "mammoth project" in a well thought-out manner, they will notice the positive effects very quickly.

Reason 4: Excessive costs and follow-up costs

High costs and follow-up costs are common problems in AI projects. There can be aspects at almost every stage of the project that result in additional costs - from unclear requirements and poor data migration to a lack of change management or insufficient consideration of scalability. An agile development process that takes changes into account and defines concrete, measurable use cases can minimize these costs. The latter then remain transparent at every stage of the project. If key framework conditions change during development, the AI process is adapted accordingly, taking into account predefined parameters such as costs or resources. Foreseeable follow-up costs can also be taken into account at the planning stage in order to ensure reliable planning even after the end of the project.

Reason 5: Ethical concerns

Last but not least, ethical concerns and the fear of job losses due to AI must be taken seriously. AI projects are therefore always also change management projects. Transparent communication that addresses concerns at an early stage and involves employees is therefore crucial. Experience has shown that innovative IT projects not only transform processes, products and structures in companies, but also the people involved. The "human factor" therefore plays a central role in all AI projects.

Starting small and agile into a digitalized future

Artificial intelligence has the potential to fundamentally change production processes in companies. The challenges are manifold, but can be mastered with a clear strategy, transparent communication and continuous training. The entry into the use of AI should be gradual and accompanied by experienced partners in order to fully exploit the potential and ensure future-proof production. Starting with an initial, concrete AI use case in the company is an agile and secure solution for entering a digitalized future with AI.

Albrecht Lottermoser, Senior Smart Factory Expert at MaibornWolff