AI as a Turnaround Lever: Four Mid-Sized Companies Show How It Really Works

7 min read

When someone mentions AI in small and medium-sized enterprises (SMEs), they often think of ChatGPT running on a company laptop. The reality is more intriguing: German SMEs are leveraging artificial intelligence to reduce production costs, better understand customers, and even revolutionize entire business models. Four companies showcase how this translates into practical applications.

Key Takeaways

Beyond the Chatbot Hype

The German AI debate is plagued by a perception issue. In the media, two extremes dominate: On one side, the dystopian view — AI eliminates jobs, replaces humans, and makes entire professions obsolete. On the other side, the Silicon Valley hype — every problem is an AI problem, and every solution requires a large language model.

The reality in the German SME sector looks different. It’s more pragmatic and results-driven. Four companies — none of which are DAX corporations, startups, or tech firms — are using AI to solve concrete business problems. Their experiences demonstrate what works, what doesn’t, and what others can learn from.

1. Schott AG: When the Machine Sees Better Than the Human Eye

Schott — the Mainz-based specialty glass manufacturer with 17,000 employees — faced a challenge as old as the glass industry itself: quality control. Each glass sheet, each pharmaceutical packaging, and each Ceran cooktop must be inspected for flaws. Traditionally, trained employees stand at the conveyor belt and visually inspect each part.

The defect rate was low — 0.3 percent. However, with millions of parts produced annually, 0.3 percent translates to thousands of faulty products slipping through the inspection process. Additionally, employees could not maintain their concentration for an entire eight-hour shift. This was both understandable and economically problematic.

Schott’s solution: a KI-powered image recognition system that analyzes each part in real-time. Trained with 2.4 million images of faulty and flawless products, the AI detects micro-level defects such as inclusions, scratches, and bubbles that are invisible to the human eye. These defects are typically only detectable under a microscope.

After six months of operation, the results were clear: the defect rate in final inspection had dropped from 0.3 percent to 0.04 percent. The inspection speed had tripled. And the employees who previously stood at the conveyor belt are now working as AI trainers, feeding the system with new defect images and optimizing recognition models.

„When someone in the SME sector talks about AI, they usually mean ChatGPT on their company laptop.“

2. Würth: The Screw King Becomes a Data Analyst

Würth — the Künzelsau-based fastening technology giant with 87,000 employees worldwide — is not a typical AI user. The company sells screws, anchors, and tools to craftsmen and industrial customers. This business has thrived for decades on personal relationships, a legendary field service team, and a catalog of 125,000 items.

The issue: With 4 million active customers and 125,000 items, it is impossible to offer the right product to each customer at the right time. The field service team can visit 20 customers per day. An algorithm, however, can analyze 4 million customers simultaneously.

Würth’s AI system analyzes order histories, industry affiliations, seasonal patterns, and external data (building permits, economic indicators, weather data) to generate an individual recommendation list for each customer. The field service team now sees on their tablet in the morning not a generic product list, but an AI-curated selection: “Customer Müller Elektrotechnik is likely to need cable binders M8 and insulation screws in the next two weeks — their last order was 43 days ago, and the typical reorder cycle is 35 days.”

The results: a 14 percent increase in first-call resolution rate, 8 percent more sales per customer visit, and a 11 percent reduction in customer churn — because the system detects early when a customer’s ordering behavior changes.

3. Welser Profile: AI in the Manufacturing Process

Welser Profile, based in the Austrian-Bavarian border region, with 2,400 employees specializing in steel, aluminum, and stainless steel profiles, faced a classic manufacturing challenge: Each profile has different specifications, and each rolling mill has unique characteristics. Optimal process parameters (temperature, speed, pressure) had to be manually determined by experienced operators.

This knowledge was stored in the minds of employees with 20 to 30 years of experience. These employees are soon to retire, creating a classic knowledge loss problem.

Welser’s approach: An AI system that analyzes 15 years of rolling production data, 2.8 billion data points, and derives optimal setup parameters for each profile geometry. The system recommends parameters to the operator, learns from corrections, and improves with each cycle.

The results: 22 percent less scrap in the initial setup. 35 percent faster setup times for new profiles. And perhaps most importantly, the ability to systematically digitize the knowledge of senior employees and make it available to the next generation.

4. The Bavarian: Insurance Becomes Personal

The Bavarian, a insurance conglomerate with 5,500 employees, is tackling a problem that affects the entire industry: customer inquiries. Claims reports, contract changes, inquiries about coverage amounts — the customer service handles thousands of repetitive and time-consuming processes daily.

The Bavarian’s AI solution goes beyond a simple chatbot. The system analyzes incoming emails and calls, automatically categorizes them, retrieves relevant contract and claim data, and generates a solution proposal for the claims adjuster. The human reviews, corrects if necessary, and sends it out — but the AI handles 80 percent of the research and formulation work.

After nine months, the average processing time per process has decreased from 14 to 4 minutes. The first-contact resolution rate (the percentage of inquiries resolved in the first interaction) has increased from 62 to 84 percent. And customer satisfaction, measured by NPS, has improved by 18 points.

Not a single job has been eliminated. Instead, the same employees handle more processes with higher quality. The freed-up time is used for complex consulting conversations — activities where human empathy and experience are irreplaceable.

What All Four Have in Common

The four examples come from different industries, vary in company size, and use AI for different purposes. However, they share four success factors:

1. AI solves a specific business problem. None of these companies implemented AI just because they had to. Each had a specific, measurable problem, and AI was the best available solution for it. Not the only one, not the coolest. The best.

2. The data was already there. Schott had millions of images. Würth had order histories. Welser had process data. The Bavarian had customer interactions. None of the companies had to first build an expensive data infrastructure. The data treasure was in the basement — AI just brought it to light.

3. Humans remain in the loop. In no case has AI replaced humans. Everywhere, humans and machines work together: AI handles the repetitive analysis, and humans make the decisions. This builds acceptance and reduces risks.

4. ROI came quickly. All four projects have paid for themselves within 4 to 14 months. The reason: The companies did not build multi-million-dollar platforms but developed targeted applications that showed results quickly. Start small, learn fast, then scale.

What Others Can Learn from This

For medium-sized enterprises planning to invest in AI by 2026, these experiences yield five practical recommendations:

• Start with the problem, not the technology. Instead of asking, “What can we do with AI?” ask, “Which of our business problems costs us the most money — and can AI solve it better than the status quo?”
• Utilize existing data. Most medium-sized enterprises sit on data treasures they have never systematically evaluated. ERP data, machine data, customer data — the foundation for AI projects is often already in place.
• Start small and fast. A pilot project with a budget of 50,000 to 100,000 Euros that delivers results in three months is more convincing than a 2-million-euro strategy that requires 18 months of lead time.
• Invest in people, not just technology. AI tools have become more affordable. What is lacking are employees who know how to implement them. Training, change management, and internal AI champions are more important than the next tool.
• Measure consistently. Define before you start which KPIs should improve — and measure them afterward. AI without success measurement is a hobby, not an investment project.

Frequently Asked Questions

Further Reading

• Edge Computing and Industry 4.0 — cloudmagazin
• AI Projects in Manufacturing — cloudmagazin
• Cloud Trends 2026 — cloudmagazin

Source of featured image: Pexels / Pixabay