The human factor: why AI projects can’t succeed without people and processes

Machines that predict when maintenance is needed. Systems that can answer complex technical questions without a specialist spending hours searching databases. Software that automatically writes code and suggests optimizations... AI offers tremendous opportunities in manufacturing.

The technology is there, but between proof-of-concept and full implementation lies a major challenge. Companies that succeed in adopting AI all share one thing: they devote just as much attention to the human and organizational aspects as to the technical side.

From data to intelligent decisions

Modern manufacturing companies collect enormous amounts of data, but the real challenge lies in turning that into actionable insights. That requires more than just smart algorithms.

Take predictive maintenance. By detecting patterns in data, algorithms can predict when a component is likely to fail. This reduces unplanned downtime and keeps production lines or logistics processes running smoothly.

Another application lies in knowledge management. When dealing with incidents involving complex machines or new hardware and software that require calibration, very specific information often needs to be retrieved from large databases. Much of that knowledge currently resides in the heads of just a few experienced specialists. A technical question might take someone twenty minutes of searching through documents and lists, a manual and iterative process. A system trained on company-specific data can answer the same questions in natural language within seconds.

AI also makes a difference in software development. Tools that assist in writing scripts, spotting bugs, or optimizing algorithms can significantly speed up developers.

Challenges in AI implementation

Many manufacturing companies have gained experience with AI pilots. These experiments are often successful, but moving toward structural implementation proves difficult. The reasons are usually organizational rather than technical.

Limited resources
AI projects require specialized expertise and time, both of which are scarce in many organizations. The shortage of IT talent is a challenge most companies recognize.

Not a top priority
AI is still relatively new. It can work, but it can also fail, and you can’t know for sure upfront. Since AI applications often focus on efficiency gains within specific teams rather than direct revenue growth, they don’t always get top priority. Convincing management of the need can be a challenge in itself.

Cultural resistance
Employees used to tried-and-true methods may see AI as a threat rather than a tool. Experienced engineers, in particular, may ask why they should change their approach if it has always worked. Building awareness and trust in new technology takes time.

The black box problem
Trust plays a crucial role in manufacturing. AI systems often work like a black box: you input data and receive an outcome, but it’s not always clear how that outcome was generated. People are cautious about leaving critical decisions about production or machinery to AI. If something goes wrong and a machine stops, the impact ripples through the entire supply chain, right down to end customers. With company reputation on the line, decision-makers tend to be cautious.

On top of that, organizational challenges arise. AI projects require collaboration between IT, operations, data analysts, and management. This calls for new processes and clear agreements about responsibilities. For traditional organizations with strict hierarchies and siloed departments, that represents a significant shift.

Success starts with people

Successful AI implementation starts with people. It’s about building trust and raising awareness, not through a one-off presentation, but through patience and persistence. These actions can help turn AI into a success:

• Appoint AI ambassadors: Enthusiastic early adopters who experiment with the technology and guide their colleagues. They can demonstrate what is, and isn’t, possible, preventing unrealistic expectations.
• Offer practical training: Not everyone needs to become a data scientist, but having a basic understanding of how machine learning works helps. Employees who know what AI can and cannot do are better equipped to use it effectively.
• Start small: Instead of overhauling the entire production process at once, begin with simple applications that deliver quick wins. For example, an internal chatbot to answer frequently asked technical questions, or software that helps generate test reports.

It’s also important to be transparent about limitations. AI won’t solve every problem and can make mistakes. By being upfront about this, you prevent disappointment and maintain trust within your team.

Governance and data quality

Alongside the human factor, AI adoption requires adapted processes. Who is responsible if an algorithm makes the wrong decision about a production setting? How do you ensure that the data the system is trained on remains up to date? And how do you safeguard privacy and security?

These questions become even more complex with new legislation such as the AI Act, as well as existing regulations like GDPR. Companies must adapt their governance to meet these requirements. It takes time and resources, but it’s unavoidable if you want to deploy AI responsibly.

And then there’s the issue of data quality. Algorithms are only as good as the information they are trained on. If your dataset is incomplete or biased, the system will suffer the consequences.

From pilot to production

Moving from proof-of-concept to real implementation requires more than just technical validation. It demands a solid implementation plan that addresses both technical and organizational aspects.

Successful companies often use a phased approach. They start with processes where mistakes don’t have major consequences and risks are low, such as optimizing maintenance schedules or automating reports. Once trust has been built, they expand into more critical applications.

AI offers tremendous opportunities. But to seize them, you need more than smart algorithms and powerful computers. It takes people who are open to change and processes designed to support them. Companies that invest in these factors create the best conditions for successful AI implementation.