With ChatGPT as an application in the field of large language models and supervised/reinforcement learning, the use and application of AI has slowly left the embryonic stage – the contours for future use are already emerging, which has resulted in a real boom and investments.
And: AI application scenarios will soon become an integral part of other applications in the area of customer relationship management, social media management or even in service such as call centers. The myriad of possible application scenarios extends from digital voice assistants (such as Google Assistant), big data analytics (e.g. for pattern recognition in user behavior), the classification of websites and advertising content according to relevance, hyper targeting and (content) personalization, chatbots in customer service, next best action or offer in marketing automation, price optimization or personalized content and content creation. In the search area, search engines such as Perplexity AI, You.com and Exa have taken off.
In reality, AI application scenarios are almost equally distributed in the areas of creation (asset creation), production (content production) and activation (e.g. next best offer, customer journey analysis; Figure 2). In other words: GenAI application scenarios currently dominate usage, while comparatively higher-value applications such as Causal AI for data analysis are only implemented in individual projects (18 percent).
The “Drama, Drama, Drama” based on Bruce Darnell However, it is already at the beginning – beyond the hype, the basics, terminology and application scenarios are still unclear. Many AI pilots have been launched – but only a few are already in an operational productive phase. The most important challenges on the company side:
- a lack of know-how (67 percent) and
- poor or non-consolidated data sets (65 percent)
- encounter a sea of currently more than 2,000 dedicated AI applications in marketing (65 percent).
The large number of available IT applications means that the functions and possible application scenarios in the field of AI can hardly be kept track of and, due to a lack of knowledge of the detailed possibilities and fit into the existing process and application landscape, a heated – albeit irrelevant – discussion about the functions and features of the respective tools arises. The accompanying interviews show that from a management perspective, this leads to concern above all about
- Data protection and data security: through disclosure of sensitive company data, which leads, among other things, to the creation of an AI Manifesto and Guidelines such as Melitta or BSH Home Appliances;
- Security risk (prompt injection): manipulating AI responses through malicious input;
- Validation and prevention of hallucinations: Ensuring the accuracy of AI-generated content and preventing “hallucinations”;
- Reputation risk: through misuse or misinterpretation of AI-generated content;
- Regulatory compliance: Comply with the evolving legal framework for the use of AI as “uncharted territory”;
- Ethical use: Consistent with company values and social norms.
New terms and mechanics are meeting a positive, hopeful crowd of (un)believing AI apostles. Case studies across various industries point the way on the “Champs-Élysées de AI”:
- Ideation Workshop: Idea generation and prioritization through different approaches such as workshops, hackathons or idea competitions such as Melitta.
- Playbook: Creation of a playbook for the development of smaller, economically viable AI applications.
- Learning Journey: systematic know-how development for further qualification. Test & Learn approach (also) in the projects.
- Diversification in large language models: use open source LLMs to avoid excessive dependence on a few providers.
- Leveraging expert knowledge: Increased connectivity and information integration by using retrieval augmented generation models to generate more accurate, contextual answers and explanations that also incorporate experiential knowledge.
- Multimodal models: Increase efficiency and effectiveness through the use of multimodal models.
- Causal AI: Launch pilot applications using Causal AI to measure the impact of interventions and make decisions based on causality rather than correlation.
- Data Management: to continue the aggregation and consolidation of our own data sets in parallel with the first AI pilot projects.
- Security and data protection: Develop guidelines as an AI manifesto for data protection, data security and ethical use and share them with other companies.
- Cross-functional and coordinated approach: concerted approach across all company levels and areas to unlock the full potential of AI and monetize it pragmatically with concrete objectives.
Dr. Ralf Strauss
Dr. Ralf Strauß is Managing Partner of MarketingTechLab GmbH, Customer Excellence GmbH, initiator of the CMO Community & Digital CMO Community, Chairman of the Board of the European Marketing Confederation (EMC) and former President of the German Marketing Association (DMV). Previously, he was Senior Vice President Digitalization Marketing & Sales in the Volkswagen Group, Global Head of Product Management CRM Marketing and CMO and Head of Corporate Development at SAP in Germany & Central Europe.
Photo: Carolin Thiersch
