01.10.2024

Harnessing the Power of AI – Part 2: Implementation

AI will bring about significant changes over the next three years – that is the view of four out of five respondents in a recent study by the consultancy firm Deloitte. Initial findings show that the impending changes are met with overwhelming approval, but also with concerns. Specifically, employees’ main concern is that they will be left behind by these developments or become redundant. Materna therefore attaches great importance to involving all departments in its AI strategy right from the start.

Adobe Stock ©

Thomas Feld
Vice President Data Economics und AI Technology Officer

Only by introducing all staff and managers to the possibilities of AI can its full potential be realised. This does not mean that the use of AI will become compulsory for everyone. An AI programme should be based on voluntary participation – just like the Materna programme, AI@Materna. It is designed as an opportunity for staff and customers to harness the possibilities of AI within their own organisation.

Integrating AI across the organisation

All departments have established AI Core Teams and Single Points of Contact (SPOCs), which take on the role of champions and developers, spreading AI knowledge across the various business units and driving its further development. Together, the AI Core Teams form the AI@Materna programme with its offerings and projects. This creates a range of AI-enhanced solutions for customers, supported by strong market partnerships.

The AI@Materna programme management coordinates cross-functional initiatives and projects and is responsible for the quality management of the AI solutions. Specialised training programmes equip staff to use AI both internally and externally. Furthermore, the entire organisation is integrating AI into day-to-day workflows to boost efficiency and optimise processes.

AI assistants

According to the industry association Bitkom, ChatGPT, MS Copilot and similar tools still face an uphill struggle in many companies. Only three per cent of companies deploy generative AI assistants on a centralised basis. There is a lack of clear strategies and usage guidelines to enable their application in day-to-day office life. This includes applications such as:

Answering questions and searching for information
Creating, summarising and translating texts
Creating design elements and images
Generating code from natural language descriptions
Structuring and developing ideas

At AI@Materna, all AI assistants have therefore been assessed for their practical suitability and safety, and guidelines for using these AI systems have been drawn up. It is important to distinguish between AI assistants that provide support based on their pre-trained ‘world knowledge’ and those that can also access internal data sources. With the launch of the web version of Microsoft Copilot Protected, all Materna staff can access the knowledge of these AI systems. Furthermore, Materna is also developing its own AI solutions based on pre-trained language models, which can also be operated on-premises or as a service to protect sensitive data.

Security and data protection guaranteed

In accordance with Microsoft’s data protection policy (Commercial Data Protection Policy), users’ inputs (prompts) in the Protected version of Microsoft Copilot are also subject to special protection. In particular, they are not used by Microsoft to train an AI, but are deleted immediately. Nevertheless, MS Copilot Web remains an external cloud service, which limits its potential uses.

For this reason, any additional assistant applications are implemented by Materna via the Materna AI Factory for clients and staff, in accordance with the relevant security requirements within the framework of AI projects. These applications rely directly on Large Language Models (LLMs) rather than on ChatGPT and similar services. This is where Aleph Alpha comes into play, for example, as it offers language models both ‘on-premises’ and ‘as a service’, enabling them to meet even the highest security requirements.

After just a few weeks, the AI assistant is already proving its worth to all staff. Public tenders can be processed and evaluated much more easily and quickly by the sales team. The marketing department also benefits from the ability to revise technical texts and generate graphics. Overall, less time is needed for online research – as well as for creating concepts of all kinds.

You can find more information in the AI white paper.

All posts from the blog series “Harnessing the Power of AI”:

Part 1 – Harnessing the Power of AI – Part 1: Strategy
Part 2 – Harnessing the Power of AI – Part 2: Implementation
Part 3 – Harnessing the Power of AI – Part 3: Application Scenarios