From the Classroom to the Platform: How AI Learning Agents for DaZ Evolved into the SteP-met Concept

Before SteP-met came along, there was a very specific problem in everyday school life: classes in which pupils were learning at very different levels – a situation that was particularly evident in the subject German as a Second Language (DaZ). Standardised worksheets rarely do justice to this situation. It was this everyday reality that gave rise to the idea for SteP-met.

The starting point: personalised learning materials

As a sixth-form teacher, I began to tailor teaching materials specifically to individual learning groups and levels – worksheets based on pupils’ actual level of learning rather than the class average.

This was time-consuming. At the same time, it clearly demonstrated the difference that individualised, context-specific support can make.

After all, the same task is not automatically the right one for every pupil. Particularly in the field of German as a second language, learning levels and individual challenges can vary considerably within a single learning group.

This practical experience gave rise to the question of how individualisation could be meaningfully developed further.

From worksheets to AI learning agents

The next step was AI-supported learning agents, initially focused in particular on German as a second language.

This revealed a fundamental point: the real added value did not lie solely in the fact that the AI could generate new content.

Rather, the crucial factor was that it could use relevant information from a specific context – such as learning progress, task type and individual challenges – as the basis for support and recommendations.

This also changed the central question.

No longer simply:

What can AI generate?

But rather:

What context does AI need to ensure that the support it provides is appropriate to the actual situation?

The insight that shaped SteP-met

This observation from the classroom became the core principle of SteP-met.

AI is particularly valuable when it does not operate ‘in isolation’, but instead accesses authorised and structured information from a specific application context in a targeted manner.

In the field of education, this may include, for example, learning progress, tasks and learning pathways. Existing learning platforms can be integrated into such an approach via LTI 1.3 or other suitable LMS interfaces.

The aim is not to replace an existing learning environment with a generic AI chatbot. Rather, the existing learning context should form the basis for personalised learning support, feedback, analyses and recommendations.

This educational approach is described in more detail on the internal ‘Solutions’ page.

Are you already using a learning platform and can you see a specific use case for AI? SteP-met can work with you to assess which existing and authorised information could be usefully utilised for a clearly defined pilot and integrated via a suitable interface.

From the classroom to other contexts

The SteP-met concept, which has a wider range of applications, gradually evolved from this pedagogical approach.

This is because the underlying principle can be applied in other contexts.

In corporate training, existing learning platforms, training data, competency models and learning pathways can form the relevant context.

In sport, for example, these include performance, match and training data. They can serve as the basis for match and performance analyses, coaching recommendations or personalised training challenges.

In corporate and customer support, on the other hand, knowledge bases, product information, internal processes and specific support and customer contexts may be relevant.

What began in the classroom as an individualised worksheet has thus become a principle applicable to a variety of fields:

Existing data and systems are linked to context-based AI via suitable interfaces.

The context changes. The principle remains the same.

Existing systems rather than building AI from scratch

This development also shapes the technological focus of SteP-met.

Institutions and companies generally already have systems, platforms and data sets in place. Rather than replacing these outright, the focus is on determining how relevant information can be extracted from them for a specific AI use case.

SteP-met is therefore not tied to a single platform or interface. Depending on the initial situation, LTI 1.3, LMS integrations, APIs, data providers or custom interfaces may be considered.

The internal ‘Integration’ page explains this technology-neutral approach in more detail.

Context-based support requires a mindful approach to data

SteP-met’s educational origins also explain why not all available data belongs in the AI context.

What matters is the information that is actually relevant and authorised for the specific task.

The architecture is designed for controlled data processing. Where the specific use case allows, information can be pseudonymised or processed within the AI context without real names. The specific technical implementation is guided by the data protection, security and governance requirements of the respective institution or company.

SteP-met relies on Microsoft Azure for the scalable operation of its systems.

Start with a clear use case

Even with new projects, the starting point remains the same as it was back in the classroom: a specific need.

That is why the focus of our collaboration is initially on clearly defined pilot projects. Such a pilot may involve a defined use case, a specified data source, a limited number of users and a clear timeframe.

This allows us to assess whether the SteP-met approach offers practical benefits in the relevant context, without turning the initial idea into an open-ended bespoke development.

SteP-met did not emerge from an abstract technological idea.

It began with a specific educational problem: providing more targeted support to pupils with different levels of learning. Individualised worksheets evolved into AI learning agents. The experience gained from these learning agents, in turn, led to an insight that today extends far beyond the classroom.

The added value of AI lies not merely in generating new content. What is crucial is the targeted use of relevant information from the actual context to provide support, analysis and recommendations.

Today, this principle is being applied to education, corporate training, sport, as well as business and customer support.

Do you have a specific use case in your school, educational institution, sports club or business?

Together, we can assess which existing data and systems can be used for a clearly defined pilot project.