‘Successful GenAI implementation is not a sprint, but a strategic transformation process’

Richard, what is GenAI and what prompted the expert report?

GenAI refers to a class of artificial intelligence (AI) models that can generate new content such as text, images and audio based on patterns they have ‘learned’ from large amounts of data they have been trained on. The potential application of these models in the banking context are diverse, but they also introduce new risks and challenges. To address this, we have collaborated with experts from various banks to compile best practices and potential risks. Based on this, we developed a framework to guide banks in implementing GenAI in a structured manner, considering strategic, organizational, and technical aspects. Our holistic approach targets not only technology experts, but also executives in business, risk and compliance.

What are the key findings of the report – and how are these reflected in the day-to-day operations of a bank like Julius Baer?

Richard: Firstly, GenAI offers great potential for increasing productivity and efficiency. Secondly, the report shows that successful implementation requires more than just technology: it is about strategic anchoring, governance, risk management and, above all, a methodical approach based on clearly defined phases. Thirdly, the risks – for example in areas like data protection or hallucinations, i.e. false or fabricated content, as well as third-party dependency – are real but manageable. And fourthly, success depends largely on whether employees are empowered and on board. GenAI is not a project for an innovation team alone, but a cultural issue.

Matthias: As we were actively involved in drafting the report, it is consistent with our experience. At the outset, it was important for us to gain our own direct experience – mostly through proof-of-concepts (PoCs) – and to define clear, measurable goals before we began implementation in a programme structure. These goals covered various dimensions such as governance, risks, technology selection and change management. Regardless of this, we consider the topic of ‘AI literacy’ to be crucial. Depending on the level of maturity and knowledge within the organisation, this topic must be given very high priority at all levels to ensure that everyone involved understands the possibilities and limitations of AI technologies and can use them effectively.

Hence, implementation is crucial. What needs to be considered here, especially in practice?

Richard: A structured approach is essential – from understanding the technology and prioritising use cases to integrating it into processes and systems. Governance, data protection and risk management must be considered from the outset. GenAI only unleashes its full potential when strategy, organisation and technology work together.

Matthias: In my view, two things are needed to be successful in the field of AI technologies: on the one hand, an interdisciplinary team with clear responsibilities to successfully master the complexity of these endeavours. On the other hand, it is important to understand that we are still at the beginning of this technology cycle. Due to the rapid pace of development, assumptions or selected technologies or providers can quickly become obsolete. Accordingly, we believe it is crucial not to develop a rigid focus on a specific technology or provider, but instead to create a platform that is as flexible as possible so that it can be adapted to changing market conditions and new technologies.

...has GenAI changed everyday work at the bank?

Matthias: We are still a long way from AI technologies changing everyday work across the entire bank. Nevertheless, many of our employees already use our internal GenAI application on a daily basis, e.g. to translate or formulate text passages. In the medium term, we see great potential in this technology, which can fundamentally change the way we work. However, successfully implementing this requires more than just providing tools: we need to continuously improve our understanding of the possibilities of GenAI – today and tomorrow – and its limitations.

So GenAI will not lead to job cuts?

Richard: The report clearly shows that GenAI does not replace people, but rather complements them – especially when it comes to repetitive, standardisable tasks. The focus is not on efficiency at any price, but on reducing workloads, ensuring quality and speeding up decision-making processes. Of course, this will change certain job profiles – but that does not automatically mean job cuts. It is crucial that banks invest early on in training, cultural change and new areas of responsibility. Those who proactively retrain and provide further education will create opportunities and dispel uncertainties.

Matthias: That is also our experience. We see GenAI as an assistance system, not a replacement. Yes, certain activities will change significantly. At the same time, entirely new requirements are emerging: prompt competence, AI-related quality assurance and an understanding of regulatory requirements for automated systems. We are therefore deliberately focusing on empowerment rather than displacement. Ultimately, it's about a new interaction between humans and machines – not an either/or situation.

How will GenAI develop further?

Richard: The report shows that the next stage could lead to ‘agentic AI,’ i.e., AI systems that plan, decide and act independently. This will increase both the potential for efficiency and the requirements for governance and control. We therefore recommend a cautious, iterative approach with clear rules, transparency mechanisms and continuous monitoring. The implementation of AI and GenAI is not a sprint, but a strategic transformation process.

Matthias: We assume that traditional GenAI models will continue to be refined and specialised. We also see the development of ‘agentic AI’ as the next big topic. Looking further into the future, we expect GenAI to merge with conventional AI applications, possibly under the term ‘neurosymbolic AI’.