The Role of Large Language Models (LLMs) in Digital Value Chains

Hardly any other technological development is currently advancing as rapidly as artificial intelligence. Whilst applications such as AI-powered search functions have quickly become part of everyday life, potential uses in a business context must be carefully examined and assessed in terms of their potential, costs and risks. One area that holds great promise for businesses of all sectors and sizes is the use of Large Language Models (LLMs) within the digital value chain. Particularly in more complex corporate structures with numerous target markets, sales channels and a diverse product range, LLMs can significantly optimise product communication processes – provided they are used correctly.

‍

1.    What are LLMs?

Large Language Models are AI models which, thanks to their neural network structure and drawing on mathematics and statistics, are capable of recognising and reproducing relationships within information, as well as generating new content.

For these LLMs to function effectively, they must be trained using large volumes of data. Through training with suitable data, fine-tuning and reinforcement learning, the model continues to learn: its predictions become more accurate, and its results and generated content gain in relevance, usability and quality. In the case of relevant business offerings, non-public company data is generally not used to train the underlying base models.

Prominent LLMs include OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude and Microsoft’s Copilot. In addition to searching for relevant information and summarising or even interpreting it, these models are also used for translation, programming and text generation tasks. An OpenAI study on the use of ChatGPT in consumer scenarios shows that 49 per cent of the messages analysed related to questions and 40 per cent to the completion of specific tasks. The remaining 11 per cent mainly concerned personal reflection, exploration and recreational use.

‍

2.    Typical use cases for LLMs in product marketing

The product content lifecycle encompasses many different tasks and process steps across various software systems, applications, business units and teams. The potential use cases for LLMs in the digital value chain are correspondingly diverse:

•    Attribute value maintenance: AI is not only helpful for data quality analysis, but can also contribute to directly resolving quality issues such as missing values by providing its own suggestions.

•    Text generation: LLMs can also support teams in creating product descriptions and marketing texts, drawing on training data that takes brand-specific phrasing into account.

•    Translation: LLMs can complement existing translation management systems and machine translation services such as DeepL. In doing so, they enable greater consideration of context, target audience, brand tone and company-specific terminology.

•    Personalisation: LLMs also show great potential in the contextual adaptation of marketing copy and product descriptions for personalised customer engagement across different channels.

•    Customer service: As chatbots, LLMs can interpret customer enquiries and provide targeted advice based on product information.

‍

3.     Business Benefits of LLMs

The use cases outlined above yield a whole range of business benefits for marketing and e-commerce teams, sales and customer service:

•    Higher data quality: What makes LLMs special is that they generate better results with the help of user feedback. Staff now only need to review inputs, rather than creating data and texts themselves. Reducing manual work also lowers the likelihood of errors, which leads to an improvement in data quality in the long term.

•    Faster time-to-market: The fewer manual steps required along the digital value chain, the shorter the process steps and entire workflows become – and the shorter the time-to-market.

•    Reduced resource consumption: This also frees up internal resources, which can then be allocated to more strategically important tasks, such as the creative development of marketing concepts or campaigns.

•    Better product experience: LLMs can not only generate generic text – they can also adapt the tone, message and value proposition to suit different target audiences and sales channels. This makes product communication more relevant to the respective target audience, which can have a positive impact on the conversion rate.

•    Greater customer loyalty: Consequently, customer loyalty can also be increased. Personalised campaigns that take into account past purchases, abandoned trolleys and browsing behaviour can enhance the relevance of customer engagement and contribute to a positive brand experience. The use of chatbots can also significantly boost customer loyalty.

•    Brand strengthening: Automating product content lifecycle management enables improved control over communication processes and more targeted marketing measures, allowing staff to focus more on the marketing strategy, the expansion of sales channels and brand building.

•    Improved competitiveness: This also goes hand in hand with improved competitiveness. When combined with analytics, shop and digital shelf data, LLMs can provide insights into how product descriptions and marketing copy can be optimised based on data, enabling companies to stand out more effectively from their competitors.

‍

4.    Best practices for using LLMs in product communication

As attractive as this potential may sound, practical experience shows that such AI projects often fail for a variety of reasons. This makes careful preparation and a well-planned approach all the more important:

•    Clear project scope: A common mistake made by companies is that AI projects are not prepared with sufficient precision. For example, it must be clearly defined which functional and non-functional requirements are to be set for and implemented by the LLM within the scope of the project. The project approach must also be defined in advance and resources planned accordingly. This includes tasks such as training the model, staff training and testing the AI.

•    Clear assessment of business value and costs: Only with such a comprehensive scope can reliable cost estimates for the project be derived. These must be weighed against the expected business benefits and quantified in cost-benefit analyses such as ROI. After all, a project does not always pay off in the long term – particularly when there are many factors to take into account.

•    Focus on data quality: The reliability of LLM applications depends largely on the quality and timeliness of the data, rules and knowledge sources used. Incomplete or contradictory information increases the risk of incorrect or ‘hallucinated’ results. Therefore, outputs should be based on verified sources, systematically validated and, in the case of critical content, approved by subject matter experts.

•    Compliance and governance: It is important to establish a governance and compliance framework for the use of LLMs at an early stage, so that data protection, information security, responsibilities and quality standards are taken into account throughout the entire lifecycle of the AI application.

•    Careful process integration: Isolated AI projects are one of the most common reasons why the potential of LLMs cannot be fully realised. LLMs must be implemented as an integral part of an end-to-end value chain so that they can help to optimise business value holistically.

‍

5.    Conclusion

LLMs have the potential to significantly optimise the digital supply chain in product communication. This applies to both quality and efficiency. This results in a whole range of business benefits that contribute to the cost-effectiveness of the relevant AI investments. However, to ensure that these business benefits can be realised on the expected scale, the right conditions must be put in place.

Speak to our AI experts and secure the maximum long-term business value from your AI projects.

‍

‍