Luxembourg, 10 April 2025 – Following an initial wave of excitement, the financial sector is taking a more grounded approach to artificial intelligence (AI): only those who set clear objectives and work with clean, structured data will realise the true potential of AI-driven tools. “We see AI as a craft – something we use to enhance efficiency and quality”, says Stephan Blohm, Board Member at securities.lu. “We apply it only where it provides genuine value.”
From streamlined reporting and more accurate risk assessments to refined trading strategies, AI has sparked high expectations in the financial industry. But the hype is giving way to reality: AI tools are not miracle solutions – they are specialised programmes with clearly defined use cases. “And they can only perform effectively after extensive training”, Blohm notes. The success of AI depends largely on the quality of the data it processes. While the foundational concepts behind machine learning date back to the 1960s, today’s computing power enables the rapid processing of massive data sets. Yet Blohm cautions, “That means nothing if I feed the system poor data or give it the wrong tasks. In those cases, I can’t expect meaningful outcomes.”
Before artificial intelligence can deliver real value, companies must first audit and prepare their own data. “Implementing AI in a corporate environment involves a lot of behind-the-scenes work”, explains Prof. Dr. Hans-Jörg von Mettenheim, Director of the Chair for Quantitative Finance and Risk Management at IPAG Business School in Paris and founder of data science firm Keynum.ai. “We need to clean and structure data before we can train AI models – or even the human teams working with them.”
Jointly with securities.lu, von Mettenheim is developing AI tools to detect inconsistencies in data and ensure regulatory details are presented consistently across documentation. AI can also improve customer-facing processes. For instance, prospective clients can be guided through onboarding steps more efficiently, thus saving time while preserving the depth of human interaction. But designing these systems is a craft in itself: it requires a deep understanding of customer behaviour, meticulous programming and continuous testing to identify and eliminate weaknesses before launch.
“If you don’t use AI, you risk inefficient processes and increased error rates”, adds Stephan Blohm, Board Member at securities.lu. “For data-driven business models, AI is an essential tool.” That tool, however, must be used responsibly. “AI should only be deployed when data protection is fully ensured,” emphasises von Mettenheim. “Sensitive data should never leave the organisation. They should be processed locally, within secure AI frameworks.”
Click here to download the german version of this article