„Our AI worked well during the significant rise in interest rates“
Artificial intelligence is finding more and more areas of application, and this now includes fund management and asset management. „Artificial intelligence and neural networks are not a new topic, but their application to the financial markets has previously failed due to computer capacities,“ explains Christian Sievers, Managing Director of Laiqon's AI subsidiary Laic, in an interview with Börsen-Zeitung. „Because financial markets in particular are extremely complex systems.“
Listed company Laiqon AG previously traded as Lloyd Fonds AG. The subsiadiary Laic is responsible for the development of artificial intelligence within the Laiqon Group. Sievers, who has a degree in business administration, trains and implements artificial intelligence, the so-called Laic Advisor, with a team of around 30 AI and machine learning experts.
Immense progress
Predicting financial markets is not easy. „Chess is easier – for example the computer Big Blue managed to beat the world chess champion Garry Kasparov in 1997,“ says Sievers. But the progress in AI has been immense, and due to increasing computing power capacity, more and more AI applications have been in use since 2011.
For many asset management models, the patterns are firmly defined. Not so with AI. „The special thing about using artificial intelligence is that no pattern is predefined, but that the neural network provides answers based on the data it finds,“ explains the AI expert.
Sievers and his team began developing their models in 2018, then went live with a BaFin licence in May 2020 and launched the first Ucits fund. Today, Laic-KI manages around 150 million euros in the three mixed funds LF-AI Defensive Multi Asset, LF-AI Balanced Multi-Asset and LF-AI Dynamic Multi-Asset, the two equity funds LF - AI Impact Equity US and LF - AI Impact Equity EU, as well as in private and institutional mandates.
AI models differ
But this is only the beginning. Several joint ventures are currently being set up with institutional partners who have secured access to the Laiqon Advisor. The Laiqon products and services are entirely digitally available, and can also be made available to third parties as white-label partners. This is expected to significantly increase the assets managed by Laic.
However, as with active managers, there are also considerable differences in AI applications. „AI models are not all the same; they are very different, which doesn't make it easy to judge the quality of an AI model,“ explains Sievers. „There are always people behind an AI model who develop it.“ In this respect, it always depends on the AI model and the developers. Against this background, different AI models are likely to lead to different results. It is, therefore, also essential to take a close look at AI.
„AI always works well where large amounts of data are available,“ explains Sievers. In this respect, small-cap investments are not necessarily the sector for the use of AI, as the database here is usually relatively small. At the macro level or with large caps, however, where a lot of data is available, it makes sense to use AI. „The majority of our data for our mixed funds is macroeconomic data and therefore corresponds to a top-down approach. The second factor is price data such as share prices,“ says Sievers. „Our database goes back to 1993.“
AI remains relatively expensive
„Setting up AI in asset management is relatively expensive,“ says the expert. Once a successful AI has been set up, the more the AI manages funds, the greater the economies of scale. In this respect, Sievers assumes that the running costs of AI-driven funds are likely to be lower than those of conventional active funds, but higher than those of purely passive ETFs that track indices.
There is also the question of how and with which approaches AI models are implemented. „We use a Bayesian approach for our AI, which goes back to a mathematician called Bayes,“ explains Sievers. „We don't work with point forecasts, but rather assign a certain probability to different forecasts.“
The AI that Sievers and his team have developed has already proven itself in difficult times. „Our AI has worked well in recent years during the significant rise in interest rates on the bond markets,“ Sievers notes. „When inflation data rose, the AI positioned itself at the short end of the bond market in particular.“ As a result, the defensive mixed fund managed by AI performed significantly better than the corresponding peer group.
