Trust through expertise: Why Netresearch relies on cutting-edge research
In today's digital world, artificial intelligence is the buzzword of the moment. AI tools are popping up everywhere, promising to make everything faster, better and smarter. But hand on heart: when it comes to your business-critical processes, your data and the trust of your customers, promises are not enough. You need results you can rely on.
What happens when an AI gets "creative" and invents facts? How do you ensure that an AI solution doesn't just reproduce data, but demonstrates real, comprehensible intelligence?
At Netresearch, we don't believe in AI as a gimmick. We believe in AI as a powerful tool that must be built on a solid, scientific foundation.
That's why we don't develop our innovative software solutions in secret. We do this in close, scientific partnership with the Web & Software Engineering (WSE) research group at Leipzig University of Applied Sciences (HTWK Leipzig ) under the direction of Prof Andreas Both.
Why a research partnership?Because trust is crucial.
Your digital projects are the result of technical expertise and diligence. When you integrate AI components - whether for data analysis, process automation or customer interaction - you need to be able to rely on two things:
- Reliability and fact-based: the AI must adhere to the content and context you specify (e.g. your company database or technical documentation) and must not invent false information.
- Adaptive intelligence: The AI must be able to understand complex contexts, adapt to the user and the situation and deliver real added value.
This is precisely where our research work comes in.
Focus 1: Making the reliability of AI measurable
One of the biggest challenges with language models (LLMs) is so-called "knowledge leakage". This happens when the AI ignores the context provided (e.g. your company data) and instead relies on its own, often generic or outdated, training knowledge.
In one of our recent publications (Gwozdz & Both, 2025), we investigated exactly that. We developed a system to systematically test how AI models react when given contradictory or incomplete information.
What does this mean for you and our solutions?
Quite simply, we understand at a scientific level when and why an AI deviates from the facts. With this knowledge, we can design and validate our AI solutions in such a way that they reliably stick to the given facts.
Focus 2: Developing intelligent and adaptive systems
A chatbot that only searches for content is not an intelligent assistant. Real digital assistance understands how users interact and what their goal is. That's why another focus of our cooperation is on learning analytics and adaptive tutoring.
Here, we are analysing how systems can adapt to the level of knowledge and needs of their users in order to measurably increase success. (A current paper on this is currently in the scientific review process and shows the depth of this research).
From research to product: the example of ScormIQ
A perfect example of how this research feeds directly into our products is ScormIQ, our AI upgrade for existing e-learning courses.
The AI tutor in ScormIQ is not just a chatbot. It benefits directly from our research:
- It is trained for source fidelity (Focus 1) to reliably explain only the course content.
- It uses the principles of adaptive learning (focus 2) to support employees individually in their learning and to close gaps in understanding in a didactically valuable way.
Expertise that is recognised
The fact that we have chosen the right partners in the WSE Group at HTWK Leipzig is demonstrated by their scientific excellence. Recent successes such as 1st place at the TenderHack Hackathon (2024), the "Overall Winner" title at the Text2SPARQL Challenge (ESWC 2025) and Best Paper Awards (ICWE 2025 & IEEE ISI 2025) prove that our partners are among the international leaders in web and AI research.
What you get out of it: A partner you can trust
The co-operation between Netresearch and HTWK Leipzig is your guarantee that you will not receive any hype products. You get solutions that:
- Are scientifically sound: Built on the latest findings on AI security and user-centred software development.
- Work reliably: Thanks to rigorous fact-based and resilience testing.
- Continuously improving: Because we never stop researching, evaluating and improving.
