If it is true that banks are primarily IT companies, then the industry may be on the verge of a golden age. Having modernized their infrastructure with the "Greenfield" approach, many institutions are now positioned to operate amid the perpetual update cycles with an improved IT foundation. However, the credit industry is not entirely ready for the new era of artificial intelligence (AI) and data analytics.

Currently, the industry is primarily in exploration mode, testing isolated applications in the frontend, such as customer dialogue systems. However, these applications are likely to remain in a dual system with human supervision. With the ChatGPT systems provided by OpenAI, there is now an opportunity to tackle significant aspects of automation.

For example, document management can be revamped: it used to take a person three days to input a loan application into the system. Now this can be done in a few hours, and prospectively, in minutes, using prompts. Additionally, the continuously growing regulatory dataset can be managed by AI, making the work of regulators easier. The adoption of AI presents an opportunity for significant advancements in the banking industry.

Another significant lever is the quality and effort required for software development. Anyone who has witnessed an "Early Adopter" create an app within ten minutes through ChatGPT instructions, making it commercially deployable in 1.0 mode, can grasp the disruptive power of AI.

This requires, on the one hand, customization of ChatGPT toolkits within the framework of so-called customization. The prerequisite for this is to have a precise target image for the application area. On the other hand, there is a need to develop custom language models for internal data control, as well as define workflows and governance for AI deployment.

These requirements lead directly to the legal situation. In mid-year, the European Parliament adopted the Artificial Intelligence Act (AI Act). The Spanish presidency hopes to conclude the trilogue by the Christmas break. Once the legal texts are published, it will take two years for the framework to take effect. Since no one can afford to sit idle for that long, banks must continue to develop their systems in anticipation of the regulations.

In essence, the AI Act categorizes AI applications into specific risk classes, which determine the extent of legal requirements. The law is intended to apply to anyone offering a product or service based on AI. While there is already considerable criticism of the AI Data Act, some of it seems ritualistic, and overall, no regulatory disaster seems to be imminent.

The benchmark for the quality of the AI Data Act will be its impact on the competitiveness of Europe. Brussels should remember that many issues can be resolved through gentle fine-tuning. This implies that the screws should not be tightened too much, to avoid the stifling of innovations in the early stages.

To understand the significance of AI in banking, one can look at the triumph of high-frequency trading in the early 2000s. It was made possible by a new generation of high-performance computers, trading in microseconds based on algorithms and utilizing new telecommunications infrastructure (fiber optics) in trading systems such as Xetra. This marked a paradigm shift. Today, financial data providers, with tools like BloombergGPT, are already poised to monetize a competitive advantage. Banks can do the same by freeing their unstructured data from silos and tapping into new revenue sources. The significant leverage, however, lies on the cost side when processes are automated through AI, putting an end to the madness of manual post-processing.