Enterprise GPTs based on a general, modular platform

This platform is based on modular components, known as ‘skills’, which can be combined flexibly. This creates a modular system that offers businesses maximum adaptability with minimal development effort.

The core of the platform

The foundation is the general implementation of assistants on various LLMs (Large Language Models) – including open-source models that can, for example, be run in a sovereign cloud. The platform also enables integration with external services and data sources.

An example: the implementation of NVIDIA’s LLM. It comprises several key components that enhance efficiency and accuracy when processing queries. The assistants utilise the models’ language and logic capabilities to understand user queries, select the appropriate skills and apply them in a targeted manner.

Advanced features for businesses

• Governance & Security
  Role-based access controls, audit logs and the choice between a sovereign cloud and on-premises operation ensure compliance and data security.
• Skill Marketplace
  An internal marketplace allows skills, once developed, to be reused multiple times and shared within the organisation – saving time and costs.
• Monitoring & optimisation
  Analysis and monitoring functions enable usage patterns to be evaluated, bottlenecks to be identified and skills to be continuously improved.
• Integration with existing systems
  ERP (e.g. SAP), CRM (e.g. Salesforce) and collaboration tools such as Microsoft Teams can be seamlessly integrated. This enables processes such as invoice approval, HR self-service and IT support to be automated.
• Domain-Specific Fine-Tuning
  For particularly demanding use cases, the platform offers fine-tuning using proprietary data, as well as integration with Retrieval-Augmented Generation (RAG).

Examples of skills and use cases

The platform covers a wide range of applications:

• Data sources: Searching in wikis or databases
• Services: Integration with email services or ticketing systems
• AI methods: image or speech recognition

Building on this, specific assistants can be developed, such as:

• A compliance assistant for checking regulatory requirements
• HR assistant for internal staff enquiries
• IT support assistant for fault analysis and ticket creation

A practical example: a knowledge management assistant linked to a wiki and a document repository, which provides staff with answers to questions about the information stored there.

Deployment and use

The configured assistants can be made available to staff via modern chat applications such as Microsoft Copilot or ChatGPT.
Alternatively, they can be integrated into other applications via a REST interface – enabling new use cases to be implemented quickly.

Conclusion: The building block for the digital future

The Materna modular system for enterprise GPTs creates a scalable, secure and flexible platform that enables companies to

1. drastically reduce development effort,
2. configure assistants to suit individual needs, and
3. introduce new use cases quickly and sustainably.

In this way, AI transforms from an innovation project into a strategic driver of growth.

Find out more about successful AI roll-outs in our white paper.