Use of local LLMs to promote climate protection and digitalisation in SMEs
The targeted use of locally operated language models (LLMs) enables small and medium-sized enterprises (SMEs) to implement innovative and environmentally friendly solutions that strengthen climate protection and support digital transformation. A particularly forward-looking approach lies in the local fine-tuning of LLMs, which reduces dependence on large cloud providers by utilising regional IT infrastructure. The Lalamo project aims to create a sustainable and data protection-friendly digital solution for chatbots based on LLMs that supports climate targets and competitiveness.
Added value for climate protection through digitalisation
Regional operation and the targeted optimisation of LLMs offer clear advantages in terms of climate protection. Dispensing with energy-intensive, global cloud infrastructures and implementing optimised, resource-saving AI models significantly reduces energy consumption. SMEs thus benefit from sustainable digitalisation aimed at efficiency and cost reduction without compromising the confidentiality and security of their data.
Local LLMs and their potential for SME workflows
LLM applications in SMEs are diverse and offer the opportunity to reduce resource consumption and speed up processes through automation and smart solutions:
-
Support: LLMs can be used to handle simple customer queries, knowledge base creation and as chatbots to improve customer retention and efficiency.
-
AI-supported chatbots and automated knowledge databases improve customer support and reduce the workload of employees.
- Knowledge-management: The automatic structuring and preparation of internal company knowledge facilitates access to information and strengthens the knowledge base. By utilising internal data in LLMs, company knowledge is structured more efficiently and is easier for employees to access.
By entering the expertise of employees, it is retained within the company and can be accessed by future generations. -
Human-resources: From managing personnel files to carrying out the application process, LLMs can optimise various HR workflows and automate routine tasks and support the HR department.
-
Tendering/promotions: LLM can also provide support when participating in tenders and grants. Tender documents can be processed in a structured manner.
Independence from international cloud providers for compliance with data protection and climate targets
Many LLM services currently require a dependency on large international cloud providers whose locations and energy practices are not always compatible with European data protection standards and climate goals. The use of local open source models (e.g. Llama, Gemma, Mistral, etc.) enables optimal EU data protection and a high level of data and IT security. The use of small models strengthens IT sustainability.
Benchmarking and evaluation for sustainability
An essential part of the project is the comparative evaluation and benchmarking of different LLM approaches in areas such as support, human resource management and knowledge management. These studies will provide a comprehensive overview of the efficiency and sustainability of the models used and support the development of further smart solutions for climate protection in SMEs.
Data protection issues with US cloud providers
A key obstacle for SMEs when using cloud-based LLMs is their dependence on large, predominantly US providers such as Amazon Web Services, Microsoft Azure or Google Cloud. This dependency raises questions about data protection and data security, as data is often stored and processed in regions that do not comply with the strict European data protection guidelines of the GDPR.
Small open source LLM in self-hosting
The availability of open source LLMs such as Llama, Gemma and Mistral makes it possible to operate these models on your own servers and optimise them for specific tasks. Through targeted fine-tuning, these LLMs can be powerful even without a large cloud connection and are sufficient for many applications without the need for immense computing power. This not only creates a data protection-compliant solution, but also an opportunity to utilise IT resources more sustainably and cost-efficiently.
Sustainability and green IT
The sustainable use of LLMs running on local servers helps to reduce the ecological footprint. Through targeted fine-tuning, smaller models optimised for the respective application can work more efficiently, which reduces energy consumption and conserves resources. Green IT aspects are playing an increasingly important role in times of climate change and are a further advantage of regional infrastructure.
Benchmarking for SME-specific use cases
In order to find the best solution for different scenarios, comprehensive benchmarking must be carried out. This involves analysing and comparing various application scenarios (e.g. support, human resources) in terms of performance, data protection and energy consumption. A comparison with external providers such as ChatGPT from OpenAI is useful here in order to determine in which areas local LLMs are just as efficient or even superior.
Importance of compliance with authorisations and access control
The targeted control of authorisations is essential for data protection and the security of sensitive company data, especially in areas such as personnel and knowledge management. Lalamo integrates the tried-and-tested EGOCMS content management system, which has been in successful use for over 25 years and offers comprehensive functions for managing document access. EGOCMS enables detailed authorisation settings that ensure that information is only accessible to authorised persons and that confidential content is managed in compliance with the GDPR. In this way, Lalamo ensures secure, precise access control that creates trust and increases efficiency in information management.
Lighthouse effect and model character for SMEs in the region
The project has the potential to serve as a beacon for small and medium-sized enterprises (SMEs) in the region that also want to rely on data protection-friendly, sustainable and customisable IT solutions. By developing and trialling an independent infrastructure for the use of Large Language Models (LLMs), a forward-looking example is being created that illustrates how SMEs can implement powerful AI applications in an innovative way - without being dependent on large international cloud providers.
Objectives: Sustainability, data protection and security
The aim of implementing regional LLMs in SMEs is to optimise processes while always complying with European data protection and security requirements and at the same time using resources efficiently. Against this background, locally operated LLMs offer the advantage that they can directly integrate and process regional knowledge and specific company data without information leaving the company.
Summary and outlook
Locally operated, optimised LLMs for SMEs offer numerous advantages: Independence from cloud providers and the associated compliance with data protection standards, increased sustainability and the targeted improvement of workflows in various business areas. A comparative analysis of these possibilities, supported by market research expertise, could help SMEs to optimally utilise the potential of LLMs and thus lay the foundation for a data protection-compliant, sustainable and future-proof business model.