How we use AI in our systems – and the role it plays in the Operating System.
AI is not a feature. It is part of the architecture.
Many companies think of AI as a single tool: a chatbot, a text generator, an analytics dashboard. We think differently. AI is not an add-on in our systems – it is an integral part of the Operating System.
This means: AI has access to the right data, is embedded in operational processes, and delivers its results where they are needed – in the workflow, not in a separate window.
The AI technologies we use
We do not use just one AI, but the right combination – depending on what the company needs. Here’s an overview of the areas we work in:
Large Language Models (LLMs)
Large language models like GPT or Claude form the foundation for many of our AI functions: text understanding, text generation, summarization, classification, data extraction. We use LLMs where language needs to be processed, understood, or generated – from automated email processing to intelligent searching in company data.
AI agents and multi-agent systems
AI agents go beyond simple inquiries. They can independently perform tasks: query databases, analyze results, initiate processes, prepare decisions. In multi-agent systems, several specialized agents work together – each with their task, coordinated by the system.
An example: One agent monitors incoming orders. A second checks resource availability. A third suggests an optimized allocation. Together, they accomplish in seconds what would take a human hours.
RAG – Retrieval Augmented Generation
LLMs know a lot – but they don’t know your company data. RAG solves this problem: Before the AI generates an answer, it searches the relevant data sources of your company and incorporates the results into its response. This way, answers are not only linguistically sound – they are based on your actual data.
Machine Learning and forecasting models
Not all AI is based on language. Machine learning models detect patterns in historical data and derive predictions from them: What orders will come next month? Where will bottlenecks occur? Which customers are at risk of churn?
These models are trained on your data and continuously improve.
Document processing and OCR
Incoming documents – invoices, delivery notes, contracts, forms – are automatically read, classified, and processed. Not rule-based, but context-driven. The AI understands what a document is, extracts the relevant information, and routes it to the appropriate places in the system.
Language and text understanding (NLP)
Natural Language Processing enables our systems to understand free text – emails, customer inquiries, internal notes. The AI recognizes intents, extracts entities, and automatically categorizes information. This eliminates manual classification and makes unstructured data usable.
Custom fine-tuning models
When standard models are not accurate enough, we train our own models – tailored to the specific data and processes of your company. Fine-tuning means: An existing model is further trained with your data until it understands the language of your industry, your processes, and your exceptional cases.
How AI fits into the Operating System
All these technologies do not operate in isolation. They are building blocks within the Operating System – and unleash their full effect only in combination with the right architecture:
| Level 1 – Order | The data foundation is created, upon which AI can build. |
| Level 2 – Automation | AI begins to take over operational tasks: analyses, text generation, automation. |
| Level 3 – Freedom | AI agents, forecasts, learning systems. The software thinks ahead. |
Without consistent data, any AI remains ineffective. That’s why we build the structure first – and implement AI where it has the greatest leverage.
What sets us apart from others
We do not integrate AI as an experiment or pilot project. We build it as an integral part of the operational architecture. From the beginning. With a clear data basis, embedded in real processes, and with human control over critical decisions. AI is the strongest lever that mid-market companies have ever had. But only if it stands on the right foundation.