AI Knowledge Base for a Cleaning Service Company

ReinPlus GmbH is a cleaning company based in Bern that cleans office buildings, medical practices, hotels, and private homes throughout the region. With 120 employees and annual revenue of CHF 3.2 million, ReinPlus is a mid-sized provider in the Swiss market. The specialty: the company also cleans sensitive environments like operating rooms, laboratories, and historic buildings with special flooring, requiring specific expertise.

CEO Sandra Meier took over the company three years ago and quickly realized that the company's most critical asset, expertise, was documented nowhere. Experienced team leaders knew exactly which cleaning agent can be used on which floor, how sensitive surfaces in historic buildings are treated, and which hygiene regulations apply in medical facilities. But this knowledge existed exclusively in their heads. Turnover in the cleaning industry is high, averaging 25% per year at ReinPlus, and every new employee had to be trained by an experienced colleague, which took weeks.

The problem had three dimensions: knowledge dependency, quality fluctuations, and high onboarding costs. When an experienced team leader was sick or resigned, expertise was missing. A concrete example: a new employee used an acid-based cleaner on a marble floor in a historic building. Damage: CHF 4,500. The team leader who would have known that only pH-neutral agents are allowed there was at another location that day. Onboarding time for new employees averaged 3 weeks of accompaniment by an experienced colleague. Cost: approximately CHF 2,400 per new hire. At 30 new hires per year, this added up to CHF 72,000 annually.

We developed an AI knowledge base accessible through a simple smartphone app. The system was built in three steps: First, we conducted structured interviews with all eight team leaders to capture and document their expertise. Over 200 hours of experience knowledge were converted into structured instructions. Then, the AI was trained to understand questions in natural language: 'How do I clean marble in a historic building?' or 'Which product for blood stains on fabric?' immediately provides the correct instructions with safety information. As a third component, we integrated image recognition: employees can take a photo of the surface, and the AI recognizes the material and recommends the appropriate cleaning procedure. All instructions are available in multiple languages: German, French, Portuguese, and Albanian, the four most common languages in the team.

The biggest challenge was not the technology but knowledge capture. Experienced cleaners often could not articulate why they do something a certain way. It was 'just like that.' We developed an interview methodology where we accompanied team leaders during their work and documented their decisions in real-time. Image recognition was trained with over 500 photos of different surfaces and materials taken directly from customer properties. Another highlight: the multilingual interface was designed so that employees with limited German skills can use the app intuitively: large buttons, clear images, minimal text.

After 5 months, the numbers show clear improvements: onboarding time for new employees dropped from 3 weeks to 3 days, an 80% reduction. New employees can work independently from day one because the app guides them through every cleaning step. Material damage from incorrect cleaning methods decreased by 85%, from an average of CHF 800 per month to CHF 120. Monthly savings total CHF 1,800 through reduced onboarding costs, fewer material damages, and more efficient workflows. Employee satisfaction increased noticeably: several employees reported feeling more confident because they can access the right instructions at any time.

The most important insight: knowledge management is not just an IT project. It is a change management project. Initially, some team leaders feared becoming 'redundant' by documenting their knowledge. We addressed this by positioning them as 'knowledge champions' and visibly appreciating their expertise. Today, they are the strongest advocates of the system because they are less interrupted by routine questions and can focus on more demanding tasks.